Purpose
We are excited to announce the 2nd Mathematical Oncology meeting in Fall 2025 (October 28th - 31st) in Tampa, Florida, hosted by Moffitt Cancer Center. This international meeting builds on our 1st successful meeting held in Phoenix, Arizona in Spring 2023 as well as the robust legacy of growing initiatives related to the application of mathematical and computational approaches in cancer biology and clinical oncology including:
- The successful NCI-funded Cancer Systems Biology Consortium (CSBC https://csbconsortium.org) and the NCI-funded Physical Sciences in Oncology Consortium (PSOC https://physics.cancer.gov) that drove a series of NCI-funded Mathematical Oncology meetings in Berkeley (2010), Tampa (2012 and 2014), Tampa (2016) and Portland (2019)
- The establishment of the Mathematical Oncology Subgroup within the Society for Mathematical Biology in 2018 (http://mathematical-oncology.org/smb-mathonco.html)
- The launching of the "Mathematical Oncology Newsletter" (http://mathematical-oncology.org/newsletter) in 2018 which now has over 2300 subscribers!
- The launching of the "Mathematical Oncology Blog" (http://mathematical-oncology.org/blog) in 2019 which now has over 100 posts
- The launching of the "Mathematical Oncology Website" (http://mathematical-oncology.org) in 2019 as a community resource bringing together the newsletter, blog, jobs, seminars, conferences, artwork, and group leaders from around the world.
To support the growing momentum in our field of Mathematical Oncology, we feel the time is ripe to establish a new regularly recurring (annual or bi-annual) meeting that will provide an international venue for collaboration, integration, training and synergy for our exciting fusion of disciplines. We look forward to seeing you in Tampa to launch the next chapter in Mathematical Oncology as we leverage the power of mathematics to drive forward positive change for patients with cancer.
Sincerely,
Sandy & Kristin
Meeting Co-Chairs
Registration
The conference will be held at the historic Don CeSar Hotel on St. Pete Beach, Florida, hosted by the Integrated Mathematical Oncology Department at Moffitt Cancer Center. All lodging for the meeting will also be hosted at the Don CeSar hotel. Please submit your abstract below, early career scientists are particularly encouraged to submit as presenters for the meeting will be selected in part from submitted submitted abstracts. During the abstract submission process there will be an opportunity to also apply for a travel award to cover both travel and accommodation. Since the conference will run directly before the IMO workshop, there is an opportunity for participants to attend both - please indicated your interest in this during the abstract submission process.
Conference RegistrationConference registration is now open for those self-funding their attendance. Travel awardees and Moffitt employees do not need to register. The registration also gives access to specially negotiated hotel rate at the Don CeSar hotel.
Conference Co-Chairs
Tampa
Hosted in Tampa, this conference will be three full days of science through early evenings. Extended mid-day breaks strategically aligned to be able to enjoy the sun whilst not missing out on the science.
Interdisciplinary focus
Talks from mathematical, biological, clinical collaborators. All are welcome.
Lodging & Accommodations
- Coming soon
Structure of the meeting
The meeting will be a mix of didactic and abstract sessions. Didactic sessions seek to provide overviews of key areas emerging in oncology and in mathematical modeling. The contributed abstract sessions will provide an opportunity for scientists to share their active research and integrate around specific themes sharing both methods and applications. Evolutionary therapy is a major theme of the meeting but will not be the only focus, depending on abstracts submitted we envisage key themes will naturally emerge such as virtual patients, mechanistic learning, immune oncology to name a few.
Posters
Posters should 4ft long and 3ft tall - landscape orientation. Please print and bring your poster to the conference, as there will be no printing available on site.
Funding
Funding for the meeting generously provided by NCI PSOC and CSBC programs as well as the Moffitt Cancer Center.
Plenary Speakers
Information Transfer through Protein Dynamics
Abstract: Individual human cancer cells often show different responses to the same treatment. In this talk I will share the quantitative experimental and computational approaches my lab has developed for studying the fate and behavior of human cells at the single-cell level. I will focus on the tumor suppressor protein p53, a transcription factor controlling genomic integrity and the response to DNA damage. In the last several years my lab has established the dynamics of p53 (changes in its levels over time) as an important mechanism controlling gene expression and cellular outcomes. In response to double-strand DNA breaks, p53 exhibits pulses of expression that allow cells to repair the damage and resume growth. Switching these pulses into a sustained response enhances the activation of terminal fates, such as apoptosis and sentences. I will present studies from the lab demonstrating how studying p53 dynamics in response to radiation and chemotherapy in single cells can guide the design and schedule of combinatorial therapy. I will also present new findings using a combination of digestion-free mass spectrometry on intact p53 proteins and global transcriptional profiling, suggesting that p53's post-translational modification state is altered between its first and second pulses of expression, and the effects these have on gene expression programs. Such an interplay between dynamics and modification may offer a strategy for hubs proteins like p53 to temporally coordinate multiple cellular processes in cells.
From Data to Dynamics: Mechanistic Learning in Cancer Modeling
Abstract: Mathematical oncology increasingly draws on complementary modeling strategies to address the complexity of tumor dynamics and treatment response. Data-driven AI models excel at extracting patterns from complex data such as imaging, clinical, and molecular characterization, offering rapid and scalable predictions, whereas mechanistic models based on biological and physical principles provide interpretability and explanatory power even in the context of limited data. Recent developments in mechanistic learning, the integration of prior biological knowledge into (deep) learning architectures, illustrate how these two paradigms can be combined to overcome their respective limitations. In this presentation, I will review recent strategies in mechanistic learning and illustrate their application in the domain of radiotherapy. Examples include spatio-temporal modeling of brain tumor growth from longitudinal imaging and prediction of treatment response dynamics using hybrid approaches combining generative computer vision and mechanistic tumor growth models, aiming for counterfactual simulations of alternative radiotherapy schedules based on probability maps of tumor progression. The value of additional data types, including histopathology and multi-omics characterization, will be discussed in the context of biology-adaptive treatment strategies. By combining the scalability of AI with the interpretability of mechanistic models, mechanistic learning offers a pathway toward predictive and clinically useful tools.
Using mathematical models to support clinical trials of evolutionary cancer therapies
Abstract: As evolutionary cancer therapies increasingly enter clinical trials, there is an urgent need for theoretical and experimental studies to identify the patients most likely to benefit, to inform trial design, and to aid the interpretation of outcomes. I will appraise three promising treatment strategies that manipulate different aspects of intra-tumour evolution. The first strategy aims to maximize the probability that resistant cells succumb to stochastic extinction during multi-strike therapy. Whereas standard clinical practice is to wait for evidence of relapse, mathematical analysis within the framework of evolutionary rescue theory reveals that the optimal time to switch to a second treatment is when the tumour is close to its minimum size, when it is likely undetectable. Strategy two is bipolar androgen therapy, which involves cycling between extremely low and supraphysiologic levels of testosterone to steer evolutionary dynamics in castrate-resistant prostate cancer. A first mathematical model of this system suggests that the treatment schedule used in previous clinical trials can be substantially improved. For the third strategy, adaptive therapy, I will present results of a spatial stochastic model that bridges the gap between deterministic ODE systems and typically intractable agent-based simulations. This novel approach shows that the predicted benefit of adaptive therapy importantly differs between two- and three-dimensional tumours. Finally, I will share unpublished experimental data revealing how treatment-sensitive and resistant cells grow and interact during adaptive therapy using EGFR inhibitors, including precise tracking of the spatial configuration of resistant subclones within tumour spheroids using confocal microscopy.
Session 1: Tuesday, October 28
| Time | Agenda Item |
|---|---|
| 1:00-1:30pm | Welcome Address (Sandy Anderson & Kristin Swanson) |
| 1:30-2:00pm | Didactic: Lance L. Munn: "Mathematical Models of the Tumor Microenvironment: From Blood Flow to Immunotherapy"Understanding the tumor microenvironment and its interactions with the host requires approaches that bridge scales and disciplines. Experimental methods reveal key mechanisms, but mathematical and computational models allow us to integrate them, dissect their contributions, and generate testable predictions. I will present a spectrum of such models spanning biophysical, physiological, and immunological domains. At the biophysical scale, we use lattice Boltzmann and agent-based models to simulate blood and lymphatic cell transport, valve mechanics, and vessel remodeling—capturing emergent processes such as mechanobiological control loops and systemic immune cell activation. At the tissue and tumor scales, poroelastic and multiscale agent-based frameworks elucidate how solid stress, angiogenesis, and plasma leakage alter the tumor environment and affect responses to therapy. Extending these approaches, physiologically based pharmacokinetic and systems biology models integrate vascular normalization, immune activation, and metabolic conditions to predict outcomes of chemotherapy, immunotherapy, CAR T cell delivery, and cancer vaccines. |
| Digital twins | |
| 2:00-2:15pm | Maximilian Strobl: "Bridging the gap: What pre-clinical experiments can teach us about math model-guided treatment scheduling"Cancers are complex and evolving diseases. To tackle this complexity there has been growing interest in developing “digital twins” – personalized computational tumor models – to better inform when and how to treat to reduce toxicity and maximize tumor control. As this idea finds traction, the crucial question is how do we ensure efficacy and safety as we translate from bench to bedside? In this study, we test the digital twin approach to treatment scheduling in vitro, in the context of EGFR+ non-small cell lung cancer. Using fluorescent, time-lapse microscopy we characterize the evolutionary dynamics of co-cultures of Gefitinib-sensitive and paired resistant cell lines (PC9) across four different treatment schedules: i) continuous therapy, ii) intermittent therapy (on/off), iii) intermittent therapy (off/on), iv) continuous therapy at half the full dose. Our results demonstrate that both the dose and the frequency of treatment influence evolutionary dynamics. Intermittent therapy minimizes final resistant cell and total cell count after six treatment changes (18 days total), across four dose levels examined (2uM, 200nM, 100nM, 20nM Gefitinib). Moreover, the off/on intermittent schedule outperforms the on/off schedule, suggesting a role for spatial competition in suppressing resistant cells. Next, we test how well three commonly used mathematical models of sensitive-resistant dynamics can predict the observed dynamics: 1) A simple exponential model, 2) A logistic model which accounts for spatial competition, and 3) A 3-population model which includes an additional subpopulation of drug-tolerant cells in the “sensitive” population. While Models 1 and 2 can capture the dynamics under continuous treatment, the more complex Model 3 is required to predict the outcomes of intermittent treatment. Our work illustrates how in vitro experiments can support the development of digital twins, and how this process can uncover new insights into drug resistance evolution in cancer. |
| 2:15-2:30pm | Luciana Melina Luque: "From measurement to decision: a tissue-aware digital-twin platform for CAR T cell dosimetry"CAR T cell therapy is one of the most exciting advances in modern cancer treatment. In this approach, a patient’s own immune cells are reprogrammed in the laboratory with a synthetic “chimeric antigen receptor” (CAR) so that they can recognise and destroy cancer cells. However, relapse and primary resistance remain common and we do not fully understand why. Experimental systems (in vitro and in vivo) are invaluable for probing CAR T functionality and persistence, but many mechanisms and “what-if” dosing questions cannot be tested directly in the lab. Agent-based models (ABMs) complement experiments by enabling exploration of dosimetry strategies and revealing emergent behaviours, while aligning with the 3Rs (Replace, Reduce, Refine) to save time, cost and animals. ABMs nevertheless require rigorous calibration and validation; without them, models may fail to capture tumour progression and lose predictive power. Progress is further limited by the scarcity of robust, organ-resolved datasets and by implementations that are computationally intensive and hard to use in clinical workflows. To address these gaps, we are assembling an integrative platform linking dry-lab modelling, wet-lab measurement and translational read-outs. At its core, our published ABM serves as the mechanistic engine; around it, an organ-to-organ atlas provides concise, identifiable, tissue-specific priors for CAR T behaviour (initially in the B-cell acute lymphoblastic leukaemia mouse context), with targeted assays to inform calibration and validation. To make virtual-trial exploration practical and accessible, we are porting the model to a high-performance ABM backend and exposing a simple interface aimed at non-computational users. By calibrating the ABM to patient-specific measurements, we instantiate a digital twin, a mechanistic replica on which to test dose, fractionation, timing and route, turning the platform into a clinically useful decision aid and closing the loop from measurement to decision. In this talk, I will present our ABM’s current capabilities and insights, outline the platform blueprint (including the atlas and high-performance deployment), and share a clear roadmap for translating tissue-aware computational twins into practical dosing and route decisions. |
| 2:30-2:45pm | Pirmin Schlicke: "Calibrating Tumor Dynamics from Single-Timepoint Biopsy Data"Two critical metrics that describe tumor dynamics are the tumor growth rate and the invasion rate into adjacent healthy tissue (1). These properties can be characterized using reaction-diffusion models that capture the spatiotemporal evolution of tumors. Retrospective fits and evaluations have revealed a high correlation of tumor growth rate and radiation response (2). Previous predictive applications estimated the two mentioned rates from pre-treatment MRI features at multiple distinct time points and were able to forecast overall survival and radiation treatment efficacy in glioblastoma (3-5). However, clinical application has been limited, as these methods require multiple pre-treatment MRI scans, which are unavailable for most cancer types on a routine clinical basis. We propose an innovative mathematical approach analyzing routinely available digitized hematoxylin and eosin (H&E)-stained biopsy tissue slides. From these slides, we extract quantitative cell centroid point patterns and utilize Fourier transformation as well as two-point correlation functions to estimate tissue-specific tumor growth and invasion parameters (6). These parameters can be used to calibrate mathematical tumor models, that effectively enable the construction of individual (passive) digital twins from minimal and widely available routine clinical data from one single time point. We validated this framework across multiple cancer types, demonstrating predictive capability for clinically relevant outcomes, including radiotherapy response, progression-free survival, and overall survival. This method significantly reduces data requirements, enables broader clinical adoption, and thus may contribute to personalized oncology management. 1. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell 2011;144:646-74 2. Rockne R, Rockhill JK, Mrugala M, Spence AM, Kalet I, Hendrickson K, et al. Predicting the efficacy of radiotherapy in individual glioblastoma patientsin vivo:a mathematical modeling approach. Physics in Medicine and Biology 2010;55:3271-85 3. Swanson KR, Alvord EC, Jr., Murray JD. A quantitative model for differential motility of gliomas in grey and white matter. Cell Prolif 2000;33:317-29 4. Jackson PR, Juliano J, Hawkins-Daarud A, Rockne RC, Swanson KR. Patient-specific mathematical neuro-oncology: using a simple proliferation and invasion tumor model to inform clinical practice. Bull Math Biol 2015;77:846-56 5. Massey SC, White H, Whitmire P, Doyle T, Johnston SK, Singleton KW, et al. Image-based metric of invasiveness predicts response to adjuvant temozolomide for primary glioblastoma. PLoS One 2020;15:e0230492 6. Pasetto S, Montejo M, Zahid MU, Rosa M, Gatenby R, Schlicke P, et al. Calibrating tumor growth and invasion parameters with spectral spatial analysis of cancer biopsy tissues. NPJ Syst Biol Appl 2024;10:112 |
| 2:45-3:00pm | Break (15 minutes) |
| 3:00-3:30pm | Didactic: Mehdi Damaghi: "Ecology and Evolution of Breast Carcinogenesis"Cancer is a complex evolving ecosystem. In the realm of medicine, we define cancer as an ecological phenomenon starting with one rebel cell breaking free from its ecological limits and multiplies rapidly disrupting the equilibrium of resident tissue homeostasis. This will eventually lead to the extinction of other species and potential ecosystem collapse with many unpredicted variations. These ecological changes of tumor microenvironment will apply novel selection pressure on cancer cells and dictates which changes in cancer cells offer adaptive advantages. To address the significance of these evolutionary and ecological processes in cancer regarding cancer initiation, progression, and metastasis, we study how various tumors are evolving in their microenvironment from normal to precancer and cancer in clinically meaningful ways. We study how changes in microenvironment of normal, precancer, and cancer cells can change their phenotype adapting to varied microenvironment and how adaptation to it can shape the new ecosystem and evolutionary trajectory of cancer cells. This interplay between tumor cells and the microenvironment plays a fundamental role in the development of an ever-changing tumor ecosystem leading to more genotypic heterogeneity and phenotypic plasticity. We use the integration of spatial single-cell transcriptomics, proteomics, metabolomic and lipidomics, and pathomics machine learning analysis to capture the heterogeneity and plasticity of cancer cells in their natural ecological microenvironment and habitats. We discovered novel metabolic phenotypic switch in cells adapted to early acidosis in mammary ducts leading to pre-cancer and carcinogenesis. We then used these markers in our breast cancer ductal carcinoma cohort to find biomarkers for progression from precancer to cancer and upstaging of DCIS. |
| Tumor microenvironment | |
| 3:30-3:45pm | Pamela Jackson: "Sex-distinct transcriptomic signatures underlie MRI-defined edema patterns in human gliomas"Magnetic resonance imaging (MRI) is key to clinically managing brain tumor patients, however connecting the biology to imaging remains challenging. We previously developed a two-compartment model of MRI signal intensity to quantitatively estimate relative edema abundance from T2-weighted MRIs. Using this model, we previously identified the fatty acid metabolism (FAM) and oxidative phosphorylation (OxPhos) pathways as sex-distinct, with both pathways amplified for high edema in males and low edema in females. The purpose of this project was to further delineate sex-distinct biology associated with MRI-estimated brain tumor edema abundance. We analyzed 179 multiregional samples (Female: 75; Male: 104) from 55 high grade glioma patients (Female: 21; Male: 34) for bulk RNA-Seq. Patients’ pre-surgical multiparametric MRIs were preprocessed and segmented for abnormal regions and normal tissue. Utilizing the segmentations and preprocessed images we estimated the relative fractions of extracellular and intracellular space based on the edema mathematical model. Samples were characterized by their edema scores, which were analyzed using differential expression for high and low edema, gene set enrichment analysis (GSEA) using MSigDB hallmarks, and leading edge interpretation. Based on transcriptomic leading edge analyses of high and low edema samples from sex-separated cohorts, we identified common and sex-distinct enrichment of the FAM and OxPhos pathways underlying edema patterns. Of the OxPhos pathway leading edge genes, 45 were common, 36 were unique to females, and 66 were unique to males. Of the FAM pathway leading edge genes, 8 were common, 39 were unique to females, and 29 were unique to males. Notably, expression of both IDH1 and IDH2 were increased for males in regions of high edema in the OxPhos pathway. IDH3a was decreased for females in regions of low edema in the OxPhos pathway. These data suggest that there may be sex-distinct metabolism underlying MRI measurable edema formation. |
| 3:45-4:00pm | David Basanta: "The only constant is change...in the tumor microenvironment"The tumor microenvironment (TME) is a critical factor in cancer initiation, progression, and treatment resistance. However, a lack of tools to study its dynamic nature has hindered our understanding of these key interactions, which likely explains important aspects of cancer progression and provides potential therapeutic targets. Our work introduces an integrated approach combining preclinical data, spatial statistics, mathematical modeling, and evolutionary learning to study TME dynamics. Our goal is to fill these knowledge gaps and design new treatment schedules that can prevent or delay the emergence of treatment resistance in cancer. |
| 4:00-4:15pm | Vural Tagal: "Modeling karyotype-driven adaptations to metabolic restrictions predict therapeutic response and immunogenicity in cancers"Cancer cells adapt to environmental and therapeutic pressures through karyotypic plasticity, including whole genome doubling, aneuploidy, and structural genome changes. Polyploid giant cancer cells (PGCCs) represent an extreme state of this plasticity, arising through endoreplication as a common response to chemotherapy, irradiation, viral infection, hypoxia, or nutrient restriction. PGCCs act as drug-tolerant reservoirs that can later depolyploidize to repopulate tumors, underscoring their role in recurrence and therapy failure. Yet, how readily tumor cells access the PGCC state and how karyotype dynamics sculpt drug sensitivity remain poorly understood. To address this gap, we integrated (i) mathematical modeling, (ii) long-term experimental evolution experiments (LTEEs), and (iii) computational analysis into a unified framework to define karyotype evolution as a predictive biomarker of therapy response. To study karyotype evolution under metabolic stress, we developed CLONEID, an LTEE platform that combines microscopy imaging, single-cell karyotyping and a neutral-drift framework distinguishing drift from selection. Its computer-vision module classifies PGCC versus proliferative states while PCA of other morphometrics revealed stress-specific phenotypes even without whole-genome doubling or overt ploidy change. LTEEs (>6 months) under glucose deprivation, phosphate restriction or hypoxia revealed stress- and ploidy-dependent trajectories. Glucose deprivation promoted whole-genome doubling whereas phosphate restriction and hypoxia drove chromosome loss toward near-diploid states. Strikingly, our in vivo studies confirmed hypoxia-induced ploidy reduction, which was predicted by LTEE trajectories within the 36 days and recapitulated by our model. Moreover, long-term adaptations also shifted therapy response. Glucose-deprived cultures gained resistance to gemcitabine and topotecan while phosphate-deprived cultures became more sensitive to taxanes and carboplatin. Furthermore, our functional assays showed that evolved cultures enhanced T-cell activation while control condition suppressed immunogenicity through tryptophan metabolism. To complement LTEEs, we built an ODE linear chain trick (LCT) model of the cell cycle, calibrated with isogenic diploid and tetraploid TNBC lines. Our model captured ploidy-conditioned therapeutic responses: Gemcitabine, among 74 tested anticancer agents, eliminated near-diploid cells but spared tetraploids that entered PGCC state and resumed proliferation after drug withdrawal. Overall, our findings identified that whole-genome doubling, polyploidy, and PGCC formation shape both long-term evolution and acute therapy responses. Integrating LTEEs with LCT modeling successfully links ploidy, karyotypic adaptation, and drug sensitivity. This approach establishes the ground for “Drug–Karyotype” pairs as biomarkers and shows how metabolic niches sculpt KFLs, offering a path toward personalized, evolution-informed oncology. |
| 4:15-4:45pm | Coffee break (30 minutes) |
| 4:45-5:15pm | Didactic: Dominik Wodarz: "Evolution in Spatially Structured Tumors"Spatial organization is a fundamental feature of tumor architecture that profoundly shapes cellular evolution and selection. This lecture examines how spatial structure influences evolutionary dynamics, with implications for driver mutation selection, tumor progression, and therapeutic resistance. I will begin by surveying spatially explicit computational modeling approaches for cellular dynamics and evolution, emphasizing the relationship between continuous-space agent-based models and coarse-grained deme models that enable analytical tractability. Through selected studies, I will explore mutant evolution in spatially structured populations, first examining expanding populations across different dimensions, then investigating mutant invasion at steady state, which is relevant for both non-neoplastic tissue and tumors in growth plateau phases during multi-step carcinogenesis. I will close by discussing our recent finding that spatial structure can reverse conventional selection: mutants with increased reproductive output may experience negative selection when multiple kinetic parameters change simultaneously, for instance, concurrent increases in both division and death rates. This challenges the assumption that higher reproductive output confers advantage. Overall, the lecture highlights how spatial population structure can fundamentally alter evolutionary principles in unexpected ways, deepening our understanding of tumor cell emergence during progression and therapy. |
| Spatial interactions | |
| 5:15-5:30pm | Tatiana Miti: "Using Agent-Based Modeling to uncover the TNBC - lung microenvironment cross-talk during metastatic niche initiation and growth"Triple-negative breast Cancer (TNBC) disproportionately affects young women with a survival median of only 1-1.5 years after the diagnosis. The short survival time in diagnosed patients is attributed to the early occurrence of metastases, where there is evidence of lung metastases in 37% of cases. Understanding how to disrupt TNBC metastatic initiation or halt the growth of existing lung metastases is crucial for preventing premature death. Unfortunately, the lack of targeted therapies for TNBC leaves chemotherapy as the main line of treatment, which shows modest results for patients with metastases. Emerging data suggest that the JNK pathway, a stress-activated molecular pathway in TNBC cells, increases both metastatic shedding from the primary tumor and survival in the lung microenvironment. JNK+ TNBC cells communicate with macrophages, endothelial cells, and fibroblasts, forming essential cross-talks that drive metastasis initiation and growth. Because chemotherapy itself activates JNK, disrupting these cross-talks is crucial for more effective therapies. Targeting single molecular components of the JNK–macrophage–vasculature or tumor–fibroblast axes in mouse xenograft models has not eliminated metastases. Direct JNK pathway inhibition also failed to fully prevent the metastatic disease, although it suppressed tumor–fibroblast signaling, reduced extracellular matrix remodeling, enhanced chemotherapy response, and limited macrophage activation. Notably, the timing of JNK inhibition (before tumor cell injection, before chemotherapy, or concurrent with chemotherapy) produced variable outcomes in metastasis number, size, and growth. To capture metastases initiation and growth dynamics, we developed an agent-based model (ABM) of TNBC–lung microenvironment ecosystem interactions. The ABM recapitulates spatiotemporal changes in TNBC tumor cells’ proliferation/death and the growth dynamics of macrophages, endothelial cells, and fibroblasts. Model predictions of the metastatic growth were validated against spatial distributions and proliferation patterns of JNK+ cells in short-term (7–10 day) mouse xenografts and compared with longer-term experimental data. Using eco-evolutionary insights from the ABM, we optimized treatment strategies aimed at disrupting tumor–lung microenvironment interactions. Results show that therapeutic success depends strongly on the sequence and spatial context of JNK+ tumor cell–lung interactions during metastatic initiation. Our ongoing work leverages this modeling framework to identify interventions capable of systematically reducing or eradicating long-term metastatic burden. While developed for TNBC, this approach provides a broadly applicable framework for studying metastatic niche ecology and guiding treatment optimization in other cancers. |
| 5:30-5:45pm | Tracy Stepien: "Modeling Tumor-Immune Interactions in the Glioblastoma Microenvironment"Glioblastoma (GBM) is an aggressive brain tumor that is extremely fatal with no current treatment options available that can achieve remission. One potential explanation for minimally effective treatments is due to the characteristically high immune-suppressive glioma microenvironment. We develop an agent-based model to simulate the interactions of glioma cells, T cells, and myeloid-derived suppressor cells (MDSCs) and the effects of oxygen, a T cell chemoattractant, and an MDSC chemoattractant. To validate our model and quantify cell clustering patterns in GBM, we use spatial statistics comparing simulations to data extracted from cross-sectional tumor images of cellular biomarkers. |
| 7:00-9:00pm | Conference Reception & Poster Session 1
All poster presenters will present in both sessions. Odd numbered posters will present from 7-8pm and Even numbered posters will present from 8-9pm.
Poster presenters:
|
Session 2: Wednesday, October 29
| Time | Agenda Item |
|---|---|
| 8:00-8:44am | Breakfast |
| 8:45-9:45am | Plenary: Galit Lahav: "Information Transfer through Protein Dynamics"Individual human cancer cells often show different responses to the same treatment. In this talk I will share the quantitative experimental and computational approaches my lab has developed for studying the fate and behavior of human cells at the single-cell level. I will focus on the tumor suppressor protein p53, a transcription factor controlling genomic integrity and the response to DNA damage. In the last several years my lab has established the dynamics of p53 (changes in its levels over time) as an important mechanism controlling gene expression and cellular outcomes. In response to double-strand DNA breaks, p53 exhibits pulses of expression that allow cells to repair the damage and resume growth. Switching these pulses into a sustained response enhances the activation of terminal fates, such as apoptosis and sentences. I will present studies from the lab demonstrating how studying p53 dynamics in response to radiation and chemotherapy in single cells can guide the design and schedule of combinatorial therapy. I will also present new findings using a combination of digestion-free mass spectrometry on intact p53 proteins and global transcriptional profiling, suggesting that p53's post-translational modification state is altered between its first and second pulses of expression, and the effects these have on gene expression programs. Such an interplay between dynamics and modification may offer a strategy for hubs proteins like p53 to temporally coordinate multiple cellular processes in cells. |
| Predictive biomarkers | |
| 9:45-10:00am | Reshmi J. S. Patel: "Predicting response to chemoradiation for cervical cancer patients with an MRI-based mathematical model"INTRODUCTION. Chemoradiation is the standard of care for locally advanced cervical cancer (LACC). However, patients still have a 20-30% risk of residual disease after completing therapy [1]. Optimizing chemoradiation to improve the therapy outcomes necessitates early and accurate predictions of tumor response on an individual basis, as population-based methods are limited in capturing tumor heterogeneity and longitudinal changes for individual patients [2]. Our goal is to accurately predict chemoradiation response for LACC patients using a biology-based mathematical model calibrated to magnetic resonance imaging (MRI) data [3]. METHODS. The patient cohort consisted of 10 LACC patients who received five weeks of concurrent cisplatin-based chemotherapy and external beam radiation [4]. Each patient underwent MRI consisting of T2-weighted, dynamic contrast-enhanced (DCE), and diffusion-weighted (DW) sequences before (V1), after two weeks (V2), and after five weeks (V3) of concurrent chemoradiation [4]. To align the tumor regions of interest (ROIs) with the MR images, we applied rigid registrations within MRI exams and non-rigid registrations (with a rigid penalty on the tumor ROI) between exams [3]. A map of the number of tumor cells, NTC(x,t), where x is the 3D position and t is time, was calculated from the DW-MRI-derived apparent diffusion coefficient map using an established method [3]. We used a reaction-diffusion model that characterizes the change in NTC(x,t) as a function of tumor cell diffusion, proliferation, and response to treatment [3]. The effect of chemotherapy was modeled as an exponential decay in tumor cells, while cell death induced by radiation was modeled as an instantaneous reduction of the number of cells from NTC(x,t) to NTCpost(x,t) according to the linear-quadratic model [3, 5]. The proliferation and chemoradiation efficacy rates were calibrated between the NTC(x,t = V1) and NTC(x,t = V2) data using the Levenberg-Marquardt algorithm. The calibrated model was run forward to make a patient-specific prediction of NTC(x,t = V3). Prediction accuracy was evaluated by calculating the concordance correlation coefficients (CCCs) between the predicted and observed changes from V1 to V3 in total tumor cellularity and tumor volume. RESULTS AND CONCLUSION. For the study cohort, the CCC between the observed and predicted change was 0.95 for the total tumor cellularity and 0.83 for the tumor volume. These results indicate that our mathematical model—calibrated with patient-specific MRI data—can accurately predict the response of cervical cancer to chemoradiation. REFERENCES. [1] Touboul C et al. Oncologist. 2010. [2] Coveney PV et al. Philos Trans A Math Phys Eng Sci. 2016. [3] Jarrett AM et al. Nat Protoc. 2021. [4] Bowen SR et al. J Magn Reson Imaging. 2018. [5] Douglas BG et al. Radiat Res. 1976. |
| 10:00-10:15am | Daniel Glazar: "Developing a model to predict hospitalization using patient-reported outcomes in cancer patients treated with radiotherapy"Patient-reported outcomes (PRO) are measures of symptoms (e.g., difficulty swallowing, insomnia, blood in stool) that are directly reported by the patient. They are measured using questionnaires that contain individual items that pertain to particular facets of a patient’s symptom (e.g., presence/absence, frequency, intensity, interference with everyday tasks). Individual items are typically reported on an ordinal n-point Likert scale from 0 (low symptom) to n-1 (high symptom). Individual symptoms can be an indication of disease progression, treatment toxicities, or comorbidities and thus can give a wholistic measure of patient wellbeing. Here, we develop a mathematical model to leverage PROs to predict hospitalization of cancer patients treated with radiotherapy. First, in order to model the effects of radiotherapy on PROs, we introduce a novel concept of radiation exposure analogous to drug exposure. Radiation exposure is modeled using a one-compartment PK model with a constant elimination rate. Additionally, we introduce another novel concept of tumor pressure to describe the effect of tumor burden on PROs. Tumor pressure is defined simply as proportional to the tumor burden. In turn, we model tumor burden dynamics using the U-shaped Claret tumor growth inhibition (TGI) model. Individual PRO items are factored into two groups: cancer-related symptoms and treatment-related symptoms. Individual PRO item response dynamics are then modeled using an n-state continuous-time inhomogeneous Markov chain model. Cancer-related symptoms include tumor pressure as a time-varying covariate in the transition rate matrix, whereas treatment-related symptoms include radiation exposure as a time-varying covariate in the transition rate matrix. We simulate PROs every 2 weeks for 104 weeks. We then model two latent variables describing overall cancer-related symptom burden and overall treatment-related symptom burden, respectively, using the 2-parameter (i.e., difficulty, discrimination) graded response model (GRM) from item response theory (IRT). Finally, we model hospitalization events using another 2-state continuous-time inhomogeneous Markov chain model with cancer-related and treatment-related symptom burden as time-varying covariates in the transition rate matrix. Future directions include validating the developed predictive model to real world data. |
| 10:15-10:30am | Konstantinos Mamis: "Quantification of the effect of early-stage cancer on cell-free DNA levels in plasma"Cell-free DNA is a promising biomarker for cancer detection. However, both the sources of elevated cell-free DNA (cfDNA) in patients with early-stage cancer, and the mechanisms by which cfDNA is shed into, and subsequently cleared from, the circulation are still poorly understood. Leveraging a rich dataset of cfDNA in healthy individuals and early-stage cancer patients, we find a multiplicative increase in cfDNA concentration in the presence of cancer. This increase is cancer type-specific, ranging from a ~1.3-fold increase in lung cancer, to a ~12-fold increase in liver cancer, and does not originate from tumor, but from healthy tissue. Employing an additional dataset reporting the tissue of origin of cfDNA, we observe a significant increase in the correlation between cfDNA originating from leukocytes and from non-leukocyte sources in cancer patients. Introducing a mathematical model for cfDNA dynamics, we find that the observed correlation can be explained by a saturation mechanism in cfDNA clearance. Saturation in clearance implies that smaller increases in cfDNA shedding from healthy tissue due to cancer may lead to the larger observed multiplicative increases in cfDNA levels. Our findings quantify cfDNA dynamics in patients with cancer, with implications for improving the accuracy of liquid biopsies for early cancer detection. |
| 10:30-11:00pm | Break (30 minutes) |
| 11:00-11:30pm | Didactic: Russ Rockne: "Localized convolutional function regression: A machine learning framework for fluid transport modeling in mathematical oncology"The integration of machine learning with mechanistic modeling is transforming the field of mathematical oncology. This lecture introduces Localized Convolutional Function Regression (LCFR), a novel AI-driven framework for analyzing dynamic contrast-enhanced MRI (DCE-MRI) data to noninvasively quantify interstitial fluid transport in tumors. LCFR leverages weak-form regression and domain-specific basis functions to estimate spatially varying coefficients of partial differential equations governing advection-diffusion-reaction dynamics. This approach enables simultaneous measurement of perfusion, diffusion, and interstitial fluid velocity in 3D, overcoming limitations of traditional voxel-wise ODE fitting and enhancing interpretability and computational efficiency. Key topics will include: The mathematical formulation of LCFR and its connection to sparse identification of nonlinear dynamics (SINDy). Validation across in silico, in vitro, and in vivo models, including hydrogel phantoms and murine glioma. Application to clinical imaging data from glioblastoma and breast cancer patients, revealing tissue-specific differences in fluid dynamics. Implications for understanding tumor microenvironment, drug delivery, and treatment response. This lecture will provide attendees with a conceptual and practical foundation for integrating AI-based model discovery into clinical imaging workflows, offering new avenues for personalized cancer modeling and predictive analytics in oncology. |
| Heterogeneity | |
| 11:30-11:45am | Sadegh Marzban: "Quantifying fitness in clonal haematopoiesis: Logistic–Moran modeling of evolutionary dynamics and therapy response"Clonal haematopoiesis (CH) is an age-associated process in which hematopoietic stem cells (HSCs) acquire somatic mutations that may confer a selective advantage, driving clonal expansion and elevating malignancy risk. Although sequencing studies have catalogued recurrent driver mutations, the evolutionary dynamics that govern their emergence, competition, and clinical impact remain poorly understood. Current models often assume exponential growth, yet such formulations neglect homeostatic constraints and competition that necessarily limit clonal expansion. Thus, more mechanistic and predictive frameworks are needed. We propose an integrated modeling strategy that combines deterministic logistic growth with stochastic Moran process simulations to study CH progression. The Moran process and its extension capture the balance of selection and drift under fixed or variable HSC pool sizes. This aligns with adult homeostasis, but can also reflect reduced compartments under stressors such as chemotherapy. Our simulations demonstrate that cytotoxic therapy contracts the HSC pool, amplifying stochastic drift and random variation in clonal prevalence. Under these conditions, mutants with modest fitness advantages expand more readily, reflecting context-dependent evolutionary success. These predictions parallel clinical observations: in a cohort breast cancer including 171 patients with serial peripheral blood samples collected before and after treatment (chemotherapy, radiation, hormone therapy, or combined modalities), the 22 patients with positively selected CH exposed to chemotherapy or chemoradiation showed significant allelic populations contractions, suggesting that therapeutic depletion of wild-type cells is a key driver of observed fitness. On the deterministic side, quantification of the differences in fitness with this improved assumption of longitudinal variant allele frequency (VAF) trajectories from the largest, publicly available dataset of CH (Bolton et al. 2020) shows that logistic growth provides a superior fit compared to exponential models. Exponential assumptions systematically underestimate clonal fitness, while logistic models capture saturation effects and more realistic growth dynamics. By unifying stochastic (Moran) and deterministic (logistic) approaches, our framework enables more accurate inference of clonal fitness and allows classification of positively versus negatively selected clones, which directly links to patient survival outcomes. This combined modeling approach establishes a mathematically rigorous foundation for quantifying CH dynamics, bridging mechanistic predictions with clinical data. It advances our ability to identify high-risk clones, forecast progression, and ultimately improve the predictive power of CH as a biomarker for patient management. |
| 11:45-12:00am | Khola Jamshad: "Modeling Cancer Evolution: A Data-Driven Agent-Based Modeling Approach to Intratumor Heterogeneity and Drug Resistance in Melanoma"Advances in multiregion sequencing have revealed extensive intratumor heterogeneity (ITH) — the presence of genetically distinct subclones within a single tumor. These findings support a branching evolution process (BEP), in which multiple subclonal lineages evolve in parallel from a common ancestor. ITH profoundly influences tumor behavior, including hallmarks of cancer, and growth and invasion phenotypes. It also underlies the emergence of therapeutic resistance, as distinct subpopulations may harbor mutations that confer survival advantages. We have built a 2D, multi-scale, data-driven agent-based model (ABM) for BEP with Hallmark Integration (BEP-HI) that simulates melanoma evolution under the hallmarks of uncontrolled proliferation, resistance to apoptosis, immune evasion, and genetic instability. The model is calibrated for BRAF-associated superficially spreading melanoma with the inclusion of key melanoma driver genes and their related hallmarks. We find three ITH regimes: clonal sweep (single dominant clone), subclonal sweep (multiple subclones cluster together), and fractal (high heterogeneity and little spatial clustering). We use cross-PCF to quantify spatial relations and find that the most aggressive clone (highest count of driver gene mutations) closely clusters with the second most aggressive. Additionally, we observe that immune cell density is higher in hot tumors than in cold tumors, and shows opposing dynamics as base mutation rates increase. By integrating differential equation models for the mechanisms of action of targeted BRAF inhibitors and anti- PD1 immunotherapy, we aim to predict how ITH affects the onset of drug resistance, and to use a virtual patient cohort to optimize combination therapy to overcome this resistance. |
| 12:00-12:15pm | Paul Schwerd-Kleine: "Modeling tumor drug responses across cells, mice, and patients"TBD |
| 12:15-1:30pm | Lunch |
| 1:30-2:00pm | Didactic: Heiko Enderling: "Radiation radiobiology unleashed: Cross-disciplinary modeling breakthroughs"Several mathematical and statistical approaches are established in radiation oncology, including the widely used Linear Quadratic (LQ) model, Biologically Effective Dose (BED), Tumor Control Probability (TCP), and Normal Tissue Complication Probability (NTCP) models. More recently, mechanistic mathematical models (including ordinary and partial differential equations) have been developed to simulate the nonlinear dynamics during radiation therapy. Mathematical models, calibrated and validated on historic clinical data, have demonstrated remarkable success in identifying innovative treatment protocols that have been subsequently validated in clinical trials in a variety of cancers and treatments. In this lecture, I will introduce the concepts of mathematical modeling and computer simulation for radiobiology and radiation biology. I will summarize cross-disciplinary modeling breakthroughs, and use individual studies to demonstrate the power of such data-driven models to guide clinical decision making, including personalizing radiation dose, dose fractionation, and organ-at-risk sparing. I will demonstrate how such models can help develop digital twins and in silico trials to learn how to best deploy radiation to individual patients. |
| Predictive models | |
| 2:00-2:15pm | Anish K. Simhal: "CALM: Dynamic, Explainable Risk Scoring for Smoldering Multiple Myeloma Progression Using an Attention-Based Deep Survival Model""ntroduction: The clinical management of smoldering multiple myeloma (SMM), a precursor condition to multiple myeloma (MM), hinges on accurately predicting its highly variable progression to active MM. While static models provide a baseline risk score, they fail to incorporate the dynamic nature of a patient's disease trajectory. The primary mathematical challenge lies in effectively modeling irregularly sampled, variable-length clinical time-series data within a survival analysis framework. Traditional survival models, such as the Cox Proportional Hazards model, are not designed to handle sequential inputs, while standard deep learning models do not naturally account for right-censored survival data. Methods: To address this, we developed a novel deep survival learning system, CALM: Cancer AI Longitudinal Modeling. We retrospectively analyzed clinical data from 438 patients at MSKCC diagnosed with SMM between 2002 and 2019, with a median follow-up of 4.24 years and a median of 11 laboratory assessments per patient (median interval between measurements: 92 days). At the last follow-up, 185 patients (42.2%) progressed to MM, while 253 (57.8%) were censored. CALM utilizes time-series measurements of key biomarkers, including serum M-protein and free light chains, along with 35 additional laboratory values. Mechanistically, CALM leverages a padded Long Short-Term Memory (LSTM) network with an attention mechanism using a Cox Proportional Hazards loss function designed to recognize patterns and trends across sequential data. The LSTM processes each patient’s longitudinal data as a sequence, learning to identify temporal changes that may signal impending progression. To translate the model's output into a clinical tool, CALM generates personalized dashboards with risk scores predicted at 3-month intervals, using only the data available up to each time point. Results: Risk scores computed achieved a mean concordance index of 0.835 ± 0.028, demonstrating strong discriminatory performance for predicting progression to MM. This represents an 11% increase in predictive potential when compared to baseline risk assessment using the Mayo 2/20/20 criteria alone in the same cohort. Visualization of the model's learned embeddings via UMAP confirmed that it successfully separated progressor and non-progressor trajectories into distinct clusters over time. The resulting patient dashboards provide a robust and interpretable tool that visualizes how changes in a patient's biomarkers directly translate into an updated risk score. Conclusions: CALM leverages routinely collected serial laboratory data to dynamically update individualized risk predictions for SMM patients. This ‘digital twin’ approach not only provides improved risk stratification over static models but also offers clinicians interpretable feedback on which laboratory trends are driving risk, potentially informing more personalized monitoring and therapeutic decision-making in clinical practice. |
| 2:15-2:30pm | Markus Schepers: "Modeling Cell-Free DNA Fragmentation Patterns for Improved Cancer Diagnostics"Cell-free DNA (cfDNA) fragmentation encodes biological signals with non-invasive diagnostic potential in oncology. This work develops mathematical and computational models to quantify fragmentation patterns and connect them to tissue-of-origin and disease state. Building on the Markovian FRIME process (Tsui et al.), a probabilistic framework is introduced for cfDNA fragmentation, enabling simulation of fragmentation dynamics and comparison with experimentally observed distributions. In parallel, data-driven approaches expand on these probabilistic features: machine learning classifiers (support vector machines, gradient boosting) are trained on fragment size profiles, end-motif frequencies, and nucleosome-positioning signals to distinguish between malignant and non-malignant cfDNA. This combination of mechanistic modeling and classification yields interpretable yet predictive tools. Preliminary results indicate that fragmentation-derived features effectively separate cancer and non-cancer samples, while the FRIME-based framework provides insight into the underlying processes. Ongoing challenges include validating fragmentation models across cohorts and sequencing platforms, and prioritizing clinical contexts where cfDNA fragmentation analysis offers unique advantages, such as early cancer detection or minimal residual disease monitoring. Together, these approaches highlight how quantitative frameworks can enhance the diagnostic utility of cfDNA in oncology. |
| 2:30-2:45pm | Heyrim Cho: "Bayesian information-theoretic approach to determine effective scanning protocols of cancer patients"The aspect of limited temporal data is one of the many challenges when dealing with clinical data. The amount of data that can be practically collected in everyday patients during the therapy is very limited due to the financial cost and the patient’s burden. This motivates us to transfer the mathematical and computational models to meet the challenges in clinical data, before we use them to guide patient therapy via prediction. In this talk, I will discuss modeling approaches to tackle this problem. For instance, I will discuss a Bayesian information-theoretic approach to determine effective scanning protocols of cancer patients. We propose a modified mutual information function with a temporal penalty term to account for the loss of temporal data. The effectiveness of our framework is demonstrated in determining scanning scheduling for radiotherapy and androgen suppression therapy patients. |
| 2:45-3:00pm | Break (15 minutes) |
| 3:00-3:30pm | Didactic: Adam MacLean: "Non-genetic paths of tumor escape"Evolutionary therapies hold great promise to ameliorate cancer outcomes, but cells also escape treatment by non-genetic means, such as through reversible, plastic cell state transitions. Epithelial-mesenchymal transition (EMT) is an example of such plasticity: cancer cells have co-opted this developmental transition to evade treatment and to seed metastases. By characterizing the gene regulation underlying EMT, we can discover means to control the spread and progression of cancer. We have developed methods to identify marker genes of highly metastatic EMT intermediate cells via mathematical modeling constrained by single-cell RNA sequencing data. Across multiple tumor types and stimuli, we identified genes consistently upregulated in EMT intermediate states, many previously unrecognized as EMT markers. We also study the gene regulatory network dynamics of EMT to understand the impact of network logic on the transition paths across the EMT landscape. We discover that choice of logic (multiplicative vs. additive gene regulation) profoundly impacts EMT phenotypes and leads to opposing predictions regarding factors that control EMT transition paths. We show that strong inhibition of miR-200 destabilizes the epithelial state and initiates EMT for multiplicative logic, in agreement with experimental data. Using single-cell data, stochastic simulations and perturbation analysis, we show how these results can be used to design experiments to infer EMT network logic in live cells. Integration of plasticity phenotypes into models of tumor evolution will provide a richer picture of the myriad ways cancer cells attempt to evade our best defenses. |
| Cell plasticity | |
| 3:30-3:45pm | Kishore Hari: "Design Principles of Gene Regulatory Networks Underlying Phenotypic Plasticity in Cancer"Phenotypic plasticity—the ability of cells to reversibly alter their functional and morphological identity—is a central tenet of cancer progression. During metastasis, cancer cells exploit developmental programs of plasticity to overcome multiple barriers. Similarly, therapy evasion is often driven by plasticity in the form of cellular persistence, which has been proposed as a precursor to therapy resistance. This adaptive behavior is orchestrated by gene regulatory networks (GRNs), which regulate protein expression and thereby determine cell phenotypes. A detailed understanding of the design principles of GRNs is therefore essential for devising strategies to target different arms of cancer. We analyzed GRNs underlying epithelial–mesenchymal plasticity (EMP), a key axis of plasticity crucial for metastasis, to identify which network features—such as structure, kinetic parameters, or expression configurations—most strongly shape the emergent “phenotypic landscape.” Our analysis uncovered a distinctive organizational motif in EMP GRNs: mutually inhibiting “teams” of nodes, with nodes within the same team activating each other’s expression while those in opposite teams inhibiting each other. This architecture, which emerges from an abundance of cooperating positive feedback loops, governs multiple properties of the EMP landscape, including the high stability and abundance of epithelial and mesenchymal phenotypes, the plasticity of hybrid E/M states that confer metastatic fitness, and the robustness of phenotypic distributions to structural and kinetic perturbations. Importantly, these effects were largely independent of other factors, such as precise kinetic parameters. Beyond EMP and cancer, we found GRNs in various contexts having strong. The networks without teams could also be viewed as teams with “impurities,” allowing us to predict the statistical features of the corresponding phenotypic landscapes. This framework thus provides a mechanistic link between network architecture and emergent dynamics. In summary, our results establish mutually inhibiting teams of nodes as a fundamental design principle of phenotypic plasticity. Beyond advancing conceptual understanding, the team framework offers a potential translational application: predicting cellular dynamics directly from patient-derived gene expression snapshots. Such predictive power could ultimately inform the design of more effective therapeutic strategies to limit cancer progression and therapy resistance. |
| 3:45-4:00pm | Joon-Hyun Song: "Microenvironmental Fluctuations Select between Storage-Effect and Plasticity-Driven Intratumor Heterogeneity"Tumor heterogeneity critically influences tumor evolution and treatment outcomes. While increased mutational burden has traditionally been considered a significant contributor to heterogeneity, phenotypic diversity often emerges independently of genetic mutations, making phenotypic variation increasingly important. Cancer cells with identical genetic profiles can still exhibit distinct phenotypes, shaped predominantly by dynamic and diverse microenvironmental conditions. Ductal carcinoma in situ (DCIS), an early-stage breast cancer that may remain stable or progress to invasive ductal carcinoma (IDC), exemplifies such environmental variability. At this stage, cancer cells experience fluctuating microenvironmental stresses, including hypoxia, acidosis, nutrient deprivation, and hormonal changes; however, the mechanisms driving this phenotypic diversity remain poorly understood. Here, we employed an agent-based model (ABM) to simulate cancer cell evolution under different scenarios of temporal environmental fluctuations. Our simulations demonstrated that environmental variability facilitates the sustained coexistence of diverse cancer cell phenotypes through the storage effect. This ecological phenomenon describes that in fluctuating environments, the population maintains diversity by storing species that favor each condition. We constructed a simplified model in which cells had defined parameters, birth and death probabilities under two distinct environmental conditions, with inherent fitness trade-offs—optimal fitness in one environment resulted in reduced fitness in another. Cell populations, initialized with randomly generated parameters and spatial constraints, were simulated under three environmental scenarios: two constant environments and one periodically fluctuating environment. Populations in constant conditions rapidly evolved towards higher average fitness in the conditions they are in, concurrently reducing heterogeneity. In contrast, populations exposed to periodic fluctuations maintained higher levels of heterogeneity despite lower average fitness in both conditions. When these evolved populations were subsequently placed into a randomly fluctuating environment, the population that adapted to periodic fluctuations exhibited superior resilience, maintained higher heterogeneity, and demonstrated greater adaptive potential compared to populations adapted to constant conditions. Our findings suggest the storage effect can be one mechanism contributing to phenotypic heterogeneity in cancer. This offers valuable insights for developing more effective cancer development and progression models. |
| 4:00-4:15pm | Marina Pérez Aliacar: "Incorporating phenotypic plasticity into MRI-informed GBM modeling"Glioblastoma (GBM) is the most common and lethal brain cancer. It has a dismal prognosis with a 5-year survival rate of 5%. The current standard of care consists of maximal surgical resection, followed by radiotherapy with concomitant and adjuvant chemotherapy with temozolomide (TMZ), the only drug approved. Unfortunately, drug resistance is the main reason behind treatment failure and responsible for the poor prognosis of patients. Cells initially respond to treatment but they eventually change their phenotype and adapt to it. That is, they modify their behavior or phenotype in response to the changes in the environment (in this case, drug exposure). This resistance is enhanced by hypoxia, a defining feature of GBM. Therefore, it is important to study how hypoxia governs cellular adaptation in GBM, since it can improve our understanding treatment response and, eventually, GBM prognosis. In view of the above, there is a need for strategies that help us understand how resistance is triggered and how it can be prevented with the final aim of designing new successful treatments. Towards this end, mathematical models enable the systematic study of these processes and, with sufficient validation with experimental data, may serve as decision support tools in clinical practice and help clinicians to determine the best treatment strategy. However, in order for them to capture the process of resistance acquisition, they must take into account plasticity, which has been included among the cancer hallmarks as a defining characteristic of the disease due to its fundamental role in processes such as drug resistance. In this context, one of the most relevant mechanisms is the go-or-grow hypothesis, which describes the phenotypic switch between proliferative and migratory states. This dichotomy not only underlies tumor invasion but also contributes to intratumoral heterogeneity, as cells within the tumor exhibit different behaviors driven by environmental pressures such as hypoxia. In this work, we introduce the dependence of proliferation on oxygen into an image-informed mathematical model of GBM calibrated with rat MRI data, implemented in FEniCsX. By integrating this phenotypic plasticity into our model, we investigate its ability to predict tumor evolution and capture key features of GBM progression. |
| 4:15-4:45pm | Coffee break (30 minutes) |
| 4:45-5:15pm | Didactic: Kristin Swanson: "Modeling and detecting shifts in the patient tumor ecosystem"TBD |
| Treatment-induced plasticity | |
| 5:15-5:30pm | Einar Bjarki Gunnarsson: "Optimal dosing of anti-cancer treatment under drug-induced plasticity"While cancer has traditionally been considered a genetic disease, mounting evidence indicates an important role for non-genetic (epigenetic) mechanisms. Common anti-cancer drugs have recently been observed to induce the adoption of non-genetic drug-tolerant cell states, thereby accelerating the evolution of drug resistance. This confounds conventional high-dose treatment strategies aimed at maximal tumor reduction, since high doses can simultaneously promote non-genetic resistance. In this talk, we propose a mathematical model of drug-induced cell plasticity, and we use this model to study optimal drug scheduling. We show that the optimal dosing strategy steers the tumor into a fixed equilibrium composition between drug-sensitive and drug-tolerant cells, while precisely balancing the trade-off between sensitive cell kill and tolerance induction. The optimal equilibrium strategy ranges from applying a low dose continuously to applying the maximum dose intermittently, depending on the dynamics of tolerance induction. We put our findings within the context of personalized medicine, where we envision a pipeline integrating mathematical modeling with experiments involving in vitro models of patient tumors, to ultimately achieve individualized model-informed dosing strategies. |
| 5:30-5:45pm | Simon Syga: "Evolution of phenotypic plasticity leads to tumor heterogeneity with implications for therapy"Cancer is a significant global health issue, with treatment challenges arising from intratumor heterogeneity. This study examines the complex relationship between somatic evolution and phenotypic plasticity, explicitly focusing on the interplay between cell migration and proliferation [1]. We propose that evolution does not act directly on phenotypic traits, like the proliferation rate, but on the phenotypic plasticity in response to the microenvironment [2]. We study this hypothesis using a novel, spatially explicit model that tracks individual cells' phenotypic and genetic states. We assume cells change between mobile and growing states controlled by inherited and mutation-driven genotypes and the cells' microenvironment. We observe that cells at the tumor edge evolve to favor migration over proliferation and vice versa in the tumor bulk. However, this phenotypic heterogeneity can be realized by distinct regulations of the phenotypic switch, which depend on the apoptosis rate and the cells' ability to sense their environment. Emerging synthetic tumors display varying levels of heterogeneity, which we show are predictors of the cancer's recurrence time after treatment. Interestingly, higher phenotypic heterogeneity predicts poor treatment outcomes, unlike genetic heterogeneity. [1] Hatzikirou, H. et al. (2010). 'Go or Grow': the key to the emergence of invasion in tumour progression? Math. Med. Biol., 29(1), 49-65. [2] Syga S. et al. (2024) Evolution of phenotypic plasticity leads to tumor heterogeneity with implications for therapy. PLOS Comput. Biol. 20(8): e1012003. |
| 7:00-9:00pm | Conference Reception & Poster Session 2
All poster presenters will present in both sessions. Even numbered posters will present from 7-8pm and Odd numbered posters will present from 8-9pm.
Poster presenters:
|
Session 3: Thursday, October 30
| Time | Agenda Item |
|---|---|
| 8:00-8:40am | Breakfast |
| 8:45-9:45am | Plenary: Rob Noble: "Using mathematical models to support clinical trials of evolutionary cancer therapies"As evolutionary cancer therapies increasingly enter clinical trials, there is an urgent need for theoretical and experimental studies to identify the patients most likely to benefit, to inform trial design, and to aid the interpretation of outcomes. I will appraise three promising treatment strategies that manipulate different aspects of intra-tumour evolution. The first strategy aims to maximize the probability that resistant cells succumb to stochastic extinction during multi-strike therapy. Whereas standard clinical practice is to wait for evidence of relapse, mathematical analysis within the framework of evolutionary rescue theory reveals that the optimal time to switch to a second treatment is when the tumour is close to its minimum size, when it is likely undetectable. Strategy two is bipolar androgen therapy, which involves cycling between extremely low and supraphysiologic levels of testosterone to steer evolutionary dynamics in castrate-resistant prostate cancer. A first mathematical model of this system suggests that the treatment schedule used in previous clinical trials can be substantially improved. For the third strategy, adaptive therapy, I will present results of a spatial stochastic model that bridges the gap between deterministic ODE systems and typically intractable agent-based simulations. This novel approach shows that the predicted benefit of adaptive therapy importantly differs between two- and three-dimensional tumours. Finally, I will share unpublished experimental data revealing how treatment-sensitive and resistant cells grow and interact during adaptive therapy using EGFR inhibitors, including precise tracking of the spatial configuration of resistant subclones within tumour spheroids using confocal microscopy. |
| Virtual trials | |
| 9:45-10:00am | Mark Robertson-Tessi: "Applications of evolutionary therapy for individual patient care and clinical trial design"Evolutionary therapy is inherently a dynamic and proactive approach to cancer treatment. This necessitates the development and use of many components typically not used in standard of care: 1) mathematical modeling is used to capture longitudinal dynamics via mechanism; 2) clinical and pre-clinical data is needed to calibrate such models and capture heterogeneity across a population; 3) an analytical framework for implementing digital twins and/or virtual patients is needed; and 4) the methods must be feasibly translatable to the clinic (for individual patient treatment decision support) and clinical trial developers (for trial design and execution). Based upon the experience of developing these components for several clinical trials (including two ‘Evolutionary Tumor Board’ (ETB) trials, NCT04343365 and NCT06423950), we summarize results and insights gained in this space. For individual patients, over 30 patients have enrolled in the ETBs for guided decision support. In addition to offering personalized evolutionary therapy options, this research has a secondary benefit of generating new hypotheses regarding how to interpret clinical data and exploit it for clinical use. Fundamental to the ETBs is the development of digital twins calibrated from retrospective data. For trial design and execution, an integrated virtual patient framework (IVPF) is being used to predict trial outcomes as well as find optimal trial design strategies. Furthermore, the IVPF can be used as a parallel virtual trial to a real trial, leveraging deep clinical calibration to provide early outcome predictions and real time mechanistic analysis of incoming trial data. |
| 10:00-10:15am | Hannah Anderson: "A framework for developing a virtual murine cohort: applications to adoptive cell therapy in bladder cancer"OBJECTIVE: Intravesical adoptive cell therapy with tumor-infiltrating lymphocytes (ACT-TIL) increases the number of tumor-specific T cells at bladder tumor site. However, the efficacy of these T cells is limited due to immune suppression from myeloid-derived suppressor cells (MDSCs). Treatment with TIL in combination with gemcitabine, which depletes MDSCs at the tumor site, has shown efficacy in a preclinical model [1]. Our aim is to develop a virtual cohort from this data for future mathematical optimization of the treatment regimen that will enable a reduction in the number of mice experiments. METHODS: Ultrasound and histology data were collected from mice with orthotopic MB49 bladder tumors treated with OT-1 cells and gemcitabine. We developed an ODE model consisting of cancer, T cells, and MDSCs with the two treatments. Structural identifiability using the differential algebra approach identified a suitable data type for model fitting. Practical identifiability was performed to determine parameters to vary for virtual cohort sampling. Parameter distributions representative of data were produced using a data-informed error threshold for the Approximate Bayesian Computation (ABC) rejection method. RESULTS: From structural identifiability analysis, model parameters are not identifiable with respect to total tumor volume data. Instead, we found that data from each cell type (cancer cells, T cells, and MDSCs) is appropriate for model fitting. Using histology and flow cytometry data, we interpolated percentages of the three cell types in the total tumor over time and used them to modify the ultrasound data to produce volume data on each cell type. This data was used to fit the model. Practical identifiability analysis showed that the tumor growth rate (p_C) and the homeostatic native T cell population (T_0) could be identified using ultrasound data alone. Thus, these parameters were not only used to generate the virtual cohort with virtual mice selected using the ABC method, but p_C and T_0 can also be used for the future creation of digital twins using ultrasound data during experimental validation of an optimized regimen. CONCLUSION: This virtual cohort framework can be applied to other diseases beyond cancer. In the future, such cohorts can be used to determine a robust, optimized treatment regimen, suitable for most subjects, and then validated in a preclinical model. CITATIONS: [1] Bazargan et al. (2023). Frontiers in Immunology, 14, 1275375. https://doi.org/10.3389/fimmu.2023.1275375 |
| 10:15-10:30am | Nicholas Harbour: "Virtual Clinical Trials of BMP4 Differentiation Therapy: Digital Twins to Aid Successful Glioblastoma Trial Design"Glioma stem cells (GSCs) are considered a major driver of glioblastoma (GBM) progression and are highly resistant to standard cytotoxic treatments. If outcomes of GBM patients are to be improved, treatments that target GSCs are urgently needed. BMP4 has been shown to drive differentiation of GSCs, increase sensitivity to radiotherapy, slow growth and increase survival times in animal models in a subset of cases. To assess the potential of BMP4 as a differentiation therapy, we develop a mathematical model that describes the growth of a GBM tumor via a hierarchy of GSCs, progenitor cells and terminally differentiated cells. To simulate treatment, we incorporate the differential sensitivity to radiation of these three cellular subpopulations as well as the effects of BMP4 on the abundance of GSCs. We parametrize our model using experimental data from twelve patient-derived GSC lines, on which we measured response to radiotherapy and population growth with and without exposure to BMP4. Cell lines were typically more sensitive to radiotherapy after two days of BMP4 treatment but population growth can either increase or decrease after seven days of exposure to BMP4. To explain this effect in our model we introduce two sensitivities to BMP4: the first describes the effect of BMP4 on GSC self-renewal and the second describes the effect of BMP4 on the proliferative capacity of progenitor cells produced by GSCs. To identify key parameters that drive successful treatment we perform global sensitivity analysis which identifies key parameters for BMP4 efficacy including proliferation rate and self-renewal sensitivity of GSCs. We then compare two treatment schedules: a single dose of BMP4 at resection and continuous delivery of BMP4 from resection till the end of radiotherapy. Due to the short half-life of BMP4 and its synergy with radiotherapy, continuous delivery of BMP4 during radiotherapy is more effective than a single dose prior to radiotherapy. We then perform a series of virtual clinical trials, stratified by tumor proliferation rate and GSC self-renewal sensitivity, which allows us to estimate the probability of observing a successful early-phase clinical trial for various virtual patient cohorts. We find that trials that selected the subset of patients with more proliferative GBMs were more likely to lead to significant improvements in survival. Significance: Targeting glioma stem cells with BMP4 provides a novel opportunity to shift the complex cellular ecosystem of gliomas to enhance treatment efficacy. Mathematical modelling can facilitate optimal patient tumor feature selection when designing successful clinical trials. |
| 10:30-11:00pm | Break (30 minutes) |
| 11:00-11:30pm | Didactic: Jingsong Zhang: "Adaptive therapy trials for metastatic prostate cancer"Androgen deprivation therapy (ADT) with luteinizing hormone-releasing hormone analog remains the cornerstone for treating metastatic prostate cancer (mPC). Recent clinical practice has added docetaxel chemotherapy and/or the androgen receptor signaling inhibitor (ARSI, e.g., abiraterone, enzalutamide, darolutamide or apalutamide) to ADT to manage newly diagnosed mPC. This has prolonged survival but mPC remains incurable. We hypothesize the current, continuous maximum tolerated dose, treatment-until-progression, approach is not evolutionarily optimal, and outcomes can be improved by integrating evolutionary dynamics into treatment of mPC. Mathematical models like the game therapy model were applied to two adaptive therapy trials (NCT03511196, NCT02415621) with ADT and ARSI and confirmed both feasibility and improved outcomes in mPC while also building an experienced team of oncologists, mathematicians, evolutionary biologists, and cancer biologists. NCT03511196 is our adaptive therapy trial for castration sensitive mPC, which showed the feasibility of using serum total PSA and testosterone levels to decide when to stop and restart ADT and or ARSI (PMID: 36358643). This study has now a median follow-up of 6 years and we have not reached the median progression free survival on the 16 enrolled patients. Exploratory biomarker analyses demonstrate patients with high post induction PSA value, high average PSA/testosterone ratio during the induction phase and the first adaptive therapy cycle tend to develop early progression if the universal 50% PSA decline criteria is used to trigger the start of treatment break. Based on what we learned from NCT03511196, we are conducting a new adaptive therapy trial with personalized PSA thresholds to trigger treatment breaks based on mathematical modeling of individual patient’s PSA and testosterone kinetics during the initial adaptive therapy cycle (NCT06734130). This study has enrolled 11/25 castration sensitive mPC pts since its opening in January 2025. |
| Evolutionary Therapy 1 | |
| 11:30-11:45am | Franco Pradelli: "Exploiting the cost of addiction in Anaplastic Large Cell Lymphoma"Drug addiction (or dependence) to therapy in cancer describes the paradoxical condition in which a resistant tumor subpopulation becomes both resistant and dependent on treatment for survival. This phenomenon, previously observed in unresectable melanoma treated with MAPK inhibitors, reveals a reduced tumor growth rate during drug holidays — an evolutionary cost that can be strategically exploited. Such observations align with the principles of adaptive and evolutionary therapy, which aim to design treatment schedules that steer and exploit the cost of resistance to reduce its burden though optimally constructed treatment strategies. In this work, we investigate drug resistance and addiction in ALK-positive Anaplastic Large Cell Lymphoma (ALCL) exposed to various ALK inhibitors (ALKi), which are used as salvage therapy for chemotherapy-resistant disease. Using a patient-derived cell line model from treatment-naïve ALCL, we induced resistance by gradually escalating ALKi concentrations over a period of more than six months. Parallel experiments tracked the evolution of drug response through longitudinal dose-response curves (DRCs), both before and after resistance and addiction emerged. Non-monotonic DRCs confirms the existence of drug-addiction to ALKi, consistent for all 3 inhibitors tested. Guided by these data, we developed a mathematical model to capture the temporal evolution of DRCs and to test how therapy design could exploit addiction in ALCL. Calibrated on experimental measurements, the model simulated tumor growth under alternative treatment strategies. Our results reveal that while continuous therapy drives resistance and tumor regrowth within ~150 days, intermittent schedules maintain long-term control for over 300 days. Furthermore, simulations suggest an adaptive strategy: switching from continuous dosing to intermittent therapy precisely when tumor burden reaches a minimum. This approach, grounded in evolutionary principles, prevents resistant populations from dominating and leverages the fitness cost of drug addiction. Additional experiments under intermittent dosing with naïve and resistant lines (both mono and co-culture) provide a preliminary validation of the effectiveness of treatment holidays to exploit ALKi drug-addiction. Moreover, they provide an opportunity to reparametrize the model to account for the effect of tumor heterogeneity. While further experimental validation is needed to validate dynamic, adaptive strategies in ALCL, our study demonstrates that integrating mathematical modeling with empirical DRCs enables a quantitative framework for predicting the evolutionary dynamics of resistance and addiction. Importantly, these results highlight the potential of adaptive and evolutionary therapy in ALCL, where optimized treatment schedules can reduce tumor burden, delay progression, and exploit the vulnerabilities created by resistance mechanisms. |
| 11:45-12:00am | Kit Gallagher: "Predicting Treatment Outcomes from Adaptive Therapy - A New Mathematical Biomarker"Standard-of-care cancer therapy regimens are characterized by continuous treatment at the maximum tolerated dose; however, this approach often fails for metastatic cancers due to the emergence of drug resistance. An evolution-based treatment paradigm known as `Adaptive Therapy' has been proposed to counter this, dynamically adjusting treatment to control, rather than minimize, the tumor burden, thus suppressing the growth of treatment-resistant cell populations and hence delaying patient relapse. Promising clinical results in prostate cancer indicate the potential of adaptive treatment protocols, but demonstrate broad heterogeneity in patient response. This naturally leads to the question: why does this heterogeneity occur, and is a `one-size-fits-all' protocol best for patients across this spectrum of responses? Using a Lotka-Volterra model to represent the dynamics of drug-sensitive and -resistant tumor populations, we derive a predictive expression for the expected benefit from Adaptive Therapy and demonstrate that this can identify the best responders in a clinical dataset. We extend this into a trio of mathematical biomarkers that can predict the time to progression and mean daily dose under a range of Adaptive Therapy protocols, accounting for clinical limitations such as fixed time intervals between clinical appointments to compare clinically realistic treatment strategies. We show that the most favorable strategy varies between patients, and present a framework to stratify patients into different treatment arms based on their individual treatment responses. Overall, the proposed strategies offer personalized treatment schedules that consistently outperform clinical standard-of-care protocols. |
| 12:00-12:15pm | Anindita Chakrabarty: "Exploiting therapy-induced senescence as an evolutionary double-bind for preventing adaptive resistance"Treatment resistance contributes to almost ninety percent of cancer-related deaths. Intratumor heterogeneity and somatic mutability, two key traits of cancer cells, are responsible for this. Overcoming this resistance, a dynamic process, requires strategies based on evolutionary biology. The 'double bind' concept, derived from predator-prey dynamics, applies to cancer when initial therapy increases the tumor's susceptibility to a subsequent treatment. A classic example is radiotherapy, making cancer cells more responsive to immune-based therapies. By appropriately sequencing the treatments and optimizing the timing of administration, the double-blind strategy can help improve patient outcomes. Our goal is to prevent adaptive resistance by exploiting a cellular fate known as therapy-induced senescence (TIS), which is induced by a wide variety of chemotherapies in approximately 31-66% of clinical samples. TIS allows cancer cells to escape the immediate cytotoxic effects of chemotherapies and emerge as more aggressive, therapy-resistant variants by triggering a temporary growth arrest. We speculated that dependence on TIS to develop chemoresistance can be exploited as a double-bind strategy. However, the success of this strategy would rely on eradicating senescent cells within the appropriate therapeutic window, mainly because TIS is a transient event. To confirm our hypothesis, we developed models of adaptive chemoresistance by inducing TIS in triple-negative breast cancer. We identified molecules produced by senescent cells that are likely responsible for the loss of therapeutic sensitivity and used these to select appropriate senolytic agents. Because we monitored the reversibility of TIS, we finally administered the senolytics within the optimal therapeutic window to assess whether this strategy could enhance the antitumor activity of the chemotherapy. As expected, the maximum effectiveness of senolytics was achieved when they were administered during the peak of senescent cell accumulation, which occurs during the transient phase of TIS. This confirmed the dynamic vulnerability of cancer cells to senescence-targeting agents and demonstrated the feasibility of exploiting this dependence as a double-bind strategy. Since TIS is a common outcome of many anticancer treatments, this finding can be validated experimentally and modeled mathematically across various cancer types and a broad range of senotherapeutics. This approach could serve as a valuable preventative strategy against the development of adaptive resistance in clinical settings. |
| 12:15-1:30pm | Lunch |
| 1:30-2:00pm | Didactic: Joel Brown: "Theory and Models for Cancer Ecology and Evolution"Cancer exhibits ecological and evolutionary dynamics and conforms to the laws and principles of ecology and evolution. First, all populations have the capacity to grow exponentially under ideal conditions. When entirely exponential, evolution selects for the trait that maximizes exponential growth rates. Second, no population can grow exponentially forever. Models of limits to growth can be phenomenological (e.g., logistic and Gompertz) or mechanistic (e.g., consumer-resource). When strictly density-dependent, natural selection in these models will select for traits that maximize population size. Density-dependent models of population growth may include Allee effects where the per capita growth rates increase with population size at very low populations. With frequency-dependent selection (the value of a trait to an individual depends on the trait values of others) evolution can result in counter-intuitive outcomes including public goods games, spite, and tragedy of the commons. Furthermore, frequency-dependent selection can promote diversification and the coexistence of different cancer cell types – as can be modelled with Lotka-Volterra competition equations or muti-species consumer resource models. When frequency-dependence favors the rare phenotype then natural selection can promote an adaptive radiation of cancer cell types filling different ecological niches. Here we will take a walk through these key concepts and models, learning their personalities and applications. I will illustrate these principles and applications using ODEs to frame and model ecological and evolutionary dynamics. Such models can be extended to PDEs, ABMs or hybrid models such as HAL as desired. In summary, the cancer’s ecological and evolutionary context, as known or hypothesized, should drive model selection and complexity. |
| Evolutionary Therapy 2 | |
| 2:00-2:15pm | Jill Gallaher: "Adaptive therapy in advanced basal cell carcinoma driven by imaging and predictive mathematical modelling"Basal cell carcinoma (BCC) accounts for 80% of skin cancers and in most cases is very treatable. However, roughly 1-10% of BCCs are considered advanced and have infiltrated into surrounding tissue such that it is challenging to clear without causing morbidity. These advanced cancers are treated with a systemic therapy inhibiting the Hedgehog signaling pathway, which is mutated in 85% of BCCs. Whilst Hedgehog inhibitors such as vismodegib have a median progression-free survival (PFS) of approximately 1 year, they come with significant toxicity that often disrupts standard treatment scheduling, and even for those that tolerate it, there is the possibility of treatment resistance. Here we present interim results from an ongoing clinical trial (NCT05651828) where we guide treatments for BCC patients using an adaptive strategy driven by imaging and mathematical modelling. Photographic images of skin lesions over time were analyzed to develop a metric for tumor burden over time and guide treatment. This metric uses a random forest algorithm to define the lesion’s current state based the color, texture, and size of a lesion to relate to its aggressiveness. We also define a system of ordinary differential equations comprising sensitive and resistant cells along with compartments to capture the drug’s unique pharmacodynamics to make predictions for future treatments. We fit the model to each patient’s lesion dynamics and predict when to start and stop therapy, iteratively driving treatment decisions to adapt treatment scheduling (either on or off) to balance growth control with toxicity and risk of drug resistance. We find considerable heterogeneity amongst patients due to the number, size, and morphology of lesions and amongst lesions within a single patient. Drug response and clearance rate of the drug may differ between patients and lesions, so pharmacokinetics and pharmacodynamics parameters need to be carefully calibrated to make good response predictions. Importantly, we define a new strategy for multiple lesions that are responding differently to treatment within the same patient. Ultimately, by combining imaging with mathematical modelling, we can predict BCC patient growth and response dynamics to make more informed treatment decisions. |
| 2:15-2:30pm | Sandhya Prabhakaran: "Evolutionary immunotherapy in NSCLC: identifying optimal dosing strategies in adoptive cell therapies using agent-based modeling"PURPOSE: Our study examines tumor growth dynamics and immunosuppression under the presence of immune attack to identify optimal immune pulsing treatment strategies. BACKGROUND: TIL therapy is an emerging immunotherapy where activated T cells are injected into the patient. This therapy can fail due to tumor-induced immunosuppression, for example via the PDL1/PD1 axis. PDL1 expression has been studied and can increase under IFNg, released by activated T cells, but PD-L1 relaxation is not understood. We hypothesize that there may be better strategies of therapy delivery that maximize tumor kill while minimizing immune suppression. We investigate these complex PD-L1 driven dynamics through a unique combination of in vitro studies and mathematical modeling. METHODS: We have collected in vitro PD-L1 expression on NSCLC cells, which were either untreated or treated for 48 hours with high dose IFNg followed by chronic low dose IFNg for 21 days. To reinforce our in vitro findings, we developed a hybrid agent-based model (ABM). The agents in the ABM are tumor cells having variable PD-L1 expression, and immune cells that secrete IFNg and can kill tumor cells. Tumor cells can be randomly or cluster seeded. IFNg is modeled as either a binary well-mixed pulse (to simulate the in vitro experimental setup), or as a diffusible via a reaction-diffusion process (to simulate the tumor-immune cell interactions in an in vivo spatial manner). We model TIL therapy as an immune cell pulse, given at regular or irregular intervals and with different quantities of cells. RESULTS: Our ABM is calibrated to capture in vitro data dynamics. Under certain combinations of spatial configurations of the tumor and intermittent immune dosing schedules, we observe the presence of ‘sweet spots’ where tumor extinction is possible and immune exhaustion is avoided. This indicates that an appropriate pulsing of TIL therapy may lead to better overall immune efficacy than a bolus injection or continuous immunotherapy, by preventing sustained suppression that increases immune cell exhaustion, despite the potential for tumor regrowth between pulses. Our novel findings can potentially benefit clinical cancer research by giving multiple insights related to the tumor extinction, equilibrium and escape phenomena. |
| 2:30-2:45pm | Sam Ganzfried: "Computing Stackelberg Equilibrium for Cancer Treatment"Recent work by Kleshnina et al. has presented a Stackelberg evolutionary game model in which the Stackelberg equilibrium strategy for the leading player corresponds to the optimal cancer treatment. We present an approach that is able to quickly and accurately solve the model presented in that work. |
| 2:45-3:00pm | Break (15 minutes) |
| 3:00-3:30pm | Didactic: Erin George: "Rewriting the rules of recurrence: integrating evolutionary insights and mathematical modeling in ovarian cancer"High-grade serous ovarian cancer (HGSOC) remains the most lethal gynecologic malignancy. Nearly 80% of patients are diagnosed at an advanced stage, and most experience recurrence despite achieving an initial clinical response to platinum-based chemotherapy. Over the past four decades, the introduction of platinum drugs and cytoreductive surgery yielded significant survival gains, but progress has since plateaued. This stagnation reflects a central paradox: while our treatments are delivered in static, uniform schedules, cancer behaves as a dynamic, evolving ecosystem. Each cycle of therapy reshapes the tumor’s fitness landscape—selecting for resistant clones, inducing cellular dormancy, and remodeling the immune and stromal microenvironment. These eco-evolutionary processes unfold across multiple timescales that remain invisible to standard clinical endpoints. Addressing this mismatch requires a shift from treating cancer as a fixed target to treating it as an adaptive system. My clinical and translational work focuses on bridging this gap through close integration of patient data, preclinical modeling, targeted therapeutics, and mathematical oncology. We aim to leverage robust clinical data through prospective collection protocols as well as patient-derived 3D microtumors and patient-derived xenografts to map how tumors evolve under treatment, while collaborating with Moffitt’s Integrated Mathematical Oncology program to translate these findings into quantitative frameworks. By parameterizing models with longitudinal measurements—tumor volume, circulating tumor DNA (ctDNA), and molecular profiles—we can simulate tumor evolution, identify critical transitions, and explore when therapy should be switched or sequenced to prevent the emergence of resistance. This approach provides a foundation for designing adaptive and evolution-informed treatment strategies in ovarian cancer. By uniting clinical observations with mathematical modeling, we can begin to predict—not just measure—how tumors change in response to therapy. In doing so, we aim to rewrite the rules of recurrence: shifting ovarian cancer management from reactive treatment of resistance to proactive steering of tumor evolution. |
| Adaptive fitness landscapes | |
| 3:30-3:45pm | Khanh N. Dinh: "Modeling the evolution of chromosomal instability in cancer from bulk and single-cell genomic data"Over the last decade, bulk DNA-sequencing (DNA-seq) has allowed us to appreciate the sheer amount and diversity of the genomic changes associated with cancer development. More recently, advances in single-cell DNA-seq have enabled profiling of copy number aberrations (CNAs) at high resolution in thousands of cells, and has uncovered the level of intra-tumor chromosomal instability (CIN). We have developed CINner, a novel mathematical model and simulation algorithm for studying single-cell dynamics in a population of cells, incorporating clonal selection of somatic driver mutations and copy number aberrations (CNAs), and accumulation of neutral passenger mutations and CNAs. CINner is efficient for large cell populations while maintaining statistical. The mathematical model and simulation method provide an efficient framework for modeling genomic changes during tumorigenesis, and are easily adaptable to accommodate newly uncovered genomic alterration mechanisms. Our first study applies CINner to uncover the selection forces driving chromosomal CNAs, estimated from large bulk DNA-seq datasets. The inferred selection parameters can predict the prevalence of whole-genome duplication in each cancer type. Moreover, the selection parameters inferred from a pan-cancer dataset show a strong correlation with the chromosomal driver gene count and potency. Together, these observations prove the biological relevance of the selection parameters inferred from CINner. The second part of this presentation delves into the development of an inference method for single-cell DNA-seq data, which holds great potentials for more accurate parameter estimation but also poses several distinct challenges over bulk data. Based on a novel implementation of random forests within the Approximate Bayesian Computation (ABC) framework, we developed an inference method to uncover both occurance rates and selection parameters driving specific CNA mechanisms in CINner. The inference recovers the parameters of interest well, even for datasets with small sample sizes. In combination, the width and depth of these studies showcase CINner’s applicability in analyzing current and upcoming DNA-seq data, toward the goal of reconstructing tumor history and predicting patient outcome. |
| 3:45-4:00pm | Richard Beck: "Local Adaptive Mapping of Karyotype Fitness Landscapes"A central challenge in evolutionary therapy is predicting tumor adaptation, yet a detailed quantitative understanding of the evolutionary dynamics of aneuploidy remains elusive. Here we introduce ALFA-K (Adaptive Local Fitness landscapes for Aneuploid Karyotypes), a novel method that infers chromosome-level karyotype fitness landscapes from longitudinal single-cell data. ALFA-K estimates the fitness of thousands of karyotypes closely related to observed populations, enabling robust prediction of emergent karyotypes not yet experimentally detected. We validated ALFA-K’s performance using synthetic data from an agent-based model and empirical data from cell lines passaged under diverse selective pressures, including cisplatin treatment. Analysis of these landscapes reveals several principles of cancer evolution with direct implications for designing and testing evolutionary therapies: 1) Therapeutic pressures and environmental context fundamentally reshape the fitness landscape by significantly modulating the selective value of copy number alterations. 2) Chromosome mis-segregation rates can dictate clonal dominance, suggesting that therapeutically modulating chromosomal instability could be used to steer tumor evolution. 3) The fitness consequences of CNAs are highly contingent on the parental karyotype, highlighting the critical role of genomic context in predicting evolutionary trajectories. 4) Whole-genome doubling facilitates more rapid karyotypic diversification by buffering cells against deleterious CNAs, thereby opening new adaptive pathways. Ultimately, ALFA-K inferred landscapes provide a quantitative basis to forecast tumor progression and rationally design strategies to pre-empt or control therapeutic resistance. |
| 4:00-4:15pm | Noemi Andor: "Karyotype-Driven Adaptation: Linking Nutrient Deprivation and Whole-Genome Duplication to Metastatic Colonization Timelines"Metastasis is the leading cause of cancer lethality, yet current predictive models can only estimate if it will occur, not when. Forecasts often deviate with an RMSE of ~15 months, limiting clinical utility. A key weakness is reliance on gene expression, which changes over time. We hypothesize that reliable prediction of metastatic timing requires modeling a driver of long-term adaptation: chromosome mis-segregations (MS). Disseminated cancer cells must adapt to foreign organ environments, a process unfolding over months or years. Somatic copy-number alterations, the primary consequence of MS, explain up to 85% of expression variance in breast cancer, underscoring their central role in adaptation. We tested this by constructing expression-informed Karyotype Fitness Landscapes (KFLs). Starting from a tumor’s karyotype and transcriptome, we simulate MS, update expression in proportion to copy number, and assign fitness using prognostic tools such as Oncotype DX. This generates landscapes linking thousands of karyotypes to predicted fitness. Using this approach we derived KFLs for breast cancer metastases in lung and brain, then simulated evolution using an agent-based model of MS we previously developed (Beck et. al, 2025). We quantified these trajectories with summary statistics capturing magnitude and dynamics of adaptation (e.g., area under the curve and slopes). Integration with MetMap phenotypic data showed that for lung metastasis, mean slope of fitness gain correlated with metastatic penetrance (r = 0.53, p = 0.061), suggesting faster adaptation predicts higher colonization likelihood. To extend these findings, we conducted long-term evolution experiments in breast and gastric cell lines under hypoxia, phosphate and glucose deprivation, with standard media as controls. Real-time growth tracking revealed progressive increases in proliferation under stress, consistent with adaptation. Karyotyping at early and late timepoints captured evolution across lineages and resource conditions. In SNU-668 cells under glucose deprivation, our neutral MS model failed to explain evolution (Fisher’s test: p ≈ 0.057). Glucose starvation imposed strong selection, with dynamics deviating sharply from neutrality (Fisher’s test: combined p ≈ 4.9 × 10⁻⁵). Both starved replicates converged on chromosome 18 loss and 21 gain as karyotype solutions. Under hypoxia, diploid SUM-159 cells remained stable, while isogenic near-tetraploid populations showed ploidy loss. In vivo results confirmed differential metastasic potential between isogenic diploid vs. tetra-ploid SUM-159 cells with ploidy loss in the latter, consistent with metabolic stress driving karyotype evolution. Together, these results establish a framework for linking MS-driven dynamics to experimental measures of metastatic behavior. By grounding predictive modeling in karyotype evolution, this approach offers a mechanistic path toward forecasting when metastasis will occur in distinct organ environments. |
| 4:15-4:45pm | Coffee break (30 minutes) |
| 4:45-5:15pm | Didactic: Bob GatenbyTBD |
| Combination therapy | |
| 5:15-5:30pm | M A Masud: "Developing model-based pipeline for combination therapy cycling: KRAS mutated NSCLC"Following relapse after first-line therapy, sotorasib is employed as a second-line treatment for KRAS-mutated non–small cell lung cancer (NSCLC). However, resistance eventually emerges, potentially through activation of compensatory signaling pathways. Here, we evaluate the efficacy of combining sotorasib with EGFR, MEK, and CDK4/6 inhibitors in vivo using xenograft models of the H358 TNBC cell line. We have developed a mathematical model using ordinary differential equations to analyze and characterize drug response dynamics and resistance rates. To maintain parameter identifiability, we employ a tumor growth inhibition (TGI) modeling framework. Analysis based on the Bliss independence criterion indicates that these drug combinations exhibit synergistic interactions, suggesting that combination therapy can enhance the therapeutic potential of second-line regimens. We further extend our modeling approach to incorporate collateral resistance and collateral sensitivity under sequential drug administration. Our mathematical analysis demonstrates that treatment output is insensitive of treatment order in the absence of collateral effects. However, the presence of collateral sensitivity is associated with delayed progression and an increased likelihood of tumor extinction. Notably, we find that a less potent drug may warrant prolonged administration compared to a more potent drug if it confers a substantially higher degree of re-sensitization to the more potent drug. This theoretical finding emphasizes a counterintuitive role of a less potent yet sensitizing drug in combination therapy. |
| 5:30-5:45pm | Charles D Kocher: "Enumerating optimal cancer treatment schedules across patient parameter space"Background and Purpose: There has been a recent recognition that continuous chemotherapy at the maximum tolerated dose (MTD) is just one of infinite possible viable treatment schedules. How to best schedule existing cancer therapies to achieve cure (extinction therapy) or maximize progression-free survival (PFS; palliative therapy) remains an open question. There is a need to systematically enumerate the optimal strategies both for extinction and palliative therapy across all regions of drug-patient-disease parameter space. Methods: Here, we used a previously introduced minimal model of cancer treatment resistance---the GDRS model, capturing tumor (exponential) Growth, cell Death due to drug (log-kill, with no collateral resistance/sensitivity), the development of Resistance to drug, and tumor re-Sensitization to drug---to study the effects of different strategies. We sampled GDRS model space by exhaustively enumerating the possible different combinations of parameter values. When possible, we did a brute-force search for the optimal treatment strategy, simulating each possible schedule and assessing it as an extinction therapy (minimum tumor volume reached) and as a palliative therapy (time until volume doubled). When the number of drugs or treatment cycles increased such that brute-force enumeration was no longer possible, we numerically found the optimum using a simulated annealing algorithm. Results: For treatment with one drug, we were able to construct a phase diagram of the optimal strategies over GDRS parameter space, for both curative intent and maximizing PFS. We found that continuous MTD treatment is the optimal palliative strategy when re-sensitization is very weak compared to the development of resistance. It is also the optimal extinction strategy, except when re-sensitization becomes strong enough that regular treatment holidays become beneficial. At intermediate levels of re-sensitization, adaptive therapy (with treatment holidays whose length depends on tumor volume regrowth) is the optimal palliative strategy. For two drugs, when re-sensitization was strong compared to resistance, rapidly alternating the drugs was the optimal strategy for both extinction therapy and palliative therapy. At very lower re-sensitization values, a “second-strike” strategy of switching at the nadir was the optimal extinction strategy. For palliative therapy at these re-sensitization values, alternating, second-strike, and sequential MTD all performed about the same. Conclusions: By constructing a phase diagram of optimal treatment strategies, we were able to identify heuristics for when various previously proposed evolution-inspired cancer treatment schedules work best. The algorithm developed here can be applied to any patient whose GDRS parameters are known to give their personalized optimal treatment schedule. Such predictions need to be tested experimentally and clinically. |
| 7:00-9:00pm | MathOnco Conference Dinner |
Session 4: Friday, October 31
| Time | Agenda Item |
|---|---|
| 8:00-8:40am | Breakfast |
| 8:45-9:45am | Plenary: Sarah Bruningk "From Data to Dynamics: Mechanistic Learning in Cancer Modeling"Mathematical oncology increasingly draws on complementary modeling strategies to address the complexity of tumor dynamics and treatment response. Data-driven AI models excel at extracting patterns from complex data such as imaging, clinical, and molecular characterization, offering rapid and scalable predictions, whereas mechanistic models based on biological and physical principles provide interpretability and explanatory power even in the context of limited data. Recent developments in mechanistic learning, the integration of prior biological knowledge into (deep) learning architectures, illustrate how these two paradigms can be combined to overcome their respective limitations. In this presentation, I will review recent strategies in mechanistic learning and illustrate their application in the domain of radiotherapy. Examples include spatio-temporal modeling of brain tumor growth from longitudinal imaging and prediction of treatment response dynamics using hybrid approaches combining generative computer vision and mechanistic tumor growth models, aiming for counterfactual simulations of alternative radiotherapy schedules based on probability maps of tumor progression. The value of additional data types, including histopathology and multi-omics characterization, will be discussed in the context of biology-adaptive treatment strategies. By combining the scalability of AI with the interpretability of mechanistic models, mechanistic learning offers a pathway toward predictive and clinically useful tools. |
| Mechanistic learning | |
| 9:45-10:00am | François de Kermenguy: "Radiation-Induced Lymphopenia: From Mathematical Modeling Towards Mechanistic Learning""Context: Radiation-induced lymphopenia (RIL) is characterized by a decrease in the absolute lymphocyte count (ALC) following radiotherapy and is associated with a significant decline in patient outcomes [1]. In this context, several mathematical models have been developed to study RIL, many of which focus on the radiation dose received by circulating blood. In 2024, our team introduced LymphoDose [2], a novel framework that goes beyond blood dose by incorporating both lymphocyte-specific recirculation mechanisms and the dose to lymphoid organs. When applied to brain tumor irradiations, we highlighted that the dose received by circulating lymphocytes appears to be too low to fully explain the high prevalence of RIL, suggesting that other mechanisms likely play an important role. This observation has been recently supported by another independent modeling study [3]. Methods and Results: To explore this hypothesis further, 184 patients with high-grade gliomas treated with chemoradiotherapy have been retrospectively identified. Patients had an average of 7.2 ± 2.2 lymphocyte blood samples collected between baseline and one-year post-treatment. The LymphoDose framework was applied to each patient. The computed lymphocyte dose metrics, along with several relevant clinical and treatment characteristics, were used for statistical analysis. In univariate analysis, each variable was assessed for its association with ALC nadir. To address multicollinearity among the selected variables, a stepwise variable selection process using Variance Inflation Factors have been performed. Finally, a multivariate analysis revealed a moderate correlation between median lymphocyte dose and ALC nadir (p=0.06). Perspectives: These results support the idea that the direct cytotoxic effect of irradiation does not fully explain the occurrence of RIL. In our last paper [4], we critically analyze the key components, assumptions, and available data underlying RIL mathematical models published to date. We propose leveraging mechanistic learning as a promising approach integrating mechanistic modeling with data-driven techniques. In particular, by providing several concrete examples, we illustrate how mechanistic learning can be utilized to develop models that are both clinically translatable and mechanistically insightful for studying RIL. [1] de Kermenguy, F. et al. Radio-induced lymphopenia in the era of anti-cancer immunotherapy. in International Review of Cell and Molecular Biology. (2023). [2 de Kermenguy, F. et al. LymphoDose: a lymphocyte dose estimation framework - application to brain radiotherapy. Phys. Med. Biol. (2024). [3] Beekman, C. et al. Radiation-Induced Lymphopenia: In Silico Replications of Preclinical Studies Suggest Importance of Dose to Lymphoid Organs. Int. J. Radiat. Oncol. Biol. Phys. 0, (2025). [4] de Kermenguy, F. et al. Radiation-Induced Lymphopenia: From Mathematical Modeling Towards Mechanistic Learning. Int. J. Radiat. Oncol. Biol. Phys. 0, (2025). " |
| 10:00-10:15am | Kayode Olumoyin: "Modeling Adoptive Cell Therapy in Bladder Cancer Using Physics-Informed Neural Network with Adaptive Loss Weighting"Intravesical adoptive cell therapy (ACT) with tumor-infiltrating lymphocytes (TIL) holds promise for durable responses in bladder cancer by administering TILs locally, thus maximizing T cells numbers in the tumor area. However, T cells in the bladder encounter an immunosuppressive population of stromal cells such as myeloid-derived suppressor cells (MDSCs), that can weaken T cell responses. Intravesical delivery of gemcitabine acts as a local lymphodepletion agent, which preconditions the bladder microenvironment for the infused T cells. To understand the underlying biological mechanisms and optimize T cell response, we employ the physics-informed neural network (PINN), with an adaptive loss weighting following the multi-task learning of the objective function to balance the contributions of the different loss terms. Using a pre-clinical murine model, bladder tumor growth was measured via ultrasound, and mice were separated into untreated, gemcitabine only (GEM), OT-I only (OT-I), and combination (GEM + OT-I) groups. An ordinary differential equation (ODE) model of tumor cells, T cells, and MDSCs interactions under the 4 different treatments is used to study the changing interactions over time between the different cells and their response to a combination of anticancer treatments. Biological constraints are enforced on the tumor cells, T cells, and MDSCs using a two-time-point histology data. We learn possible trajectories of the different interacting cells at time points where no observed data are available. We infer time-varying interactions between cells, and their response to a combination of anticancer treatments and describe its underlying implication for the biological mechanism of T cell response. |
| 10:15-10:30am | Dhananjay Bhaskar: "STAGED: A Multi-Agent Neural Network for Learning Cellular Interaction Dynamics"Single-cell technologies have enabled major advances in identifying cellular states and subpopulations, but conventional methods often treat cells as independent points, overlooking the spatial organization and dynamic cell–cell interactions that drive biological function. Spatial transcriptomics now offers a window into these interactions, but new computational frameworks are required to model the complex interplay of intra- and intercellular dynamics. We introduce Spatio-Temporal Agent-Based Graph Evolution Dynamics (STAGED), a deep learning framework that integrates agent-based modeling with graph neural ODEs to jointly learn intracellular gene regulatory networks and intercellular communication. Each cell is modeled as an ""agent"" whose internal dynamics evolve through history-aware, attention-based regulatory graphs, while signals are exchanged between neighboring cells via ligand–receptor interactions. We demonstrate STAGED’s capabilities across simulations and real-world data. In tumor microenvironment simulations, STAGED accurately recapitulates heterogeneous population dynamics in response to an oncolytic virus. In oscillatory systems, it recovers temporally varying regulatory interactions and out-of-phase gene expression dynamics across spatially coupled cells. Finally, when applied to spatial transcriptomics data from early Alzheimer’s disease, STAGED uncovers microglial subpopulations with distinct regulatory programs, suggesting that diverse gene-gene interaction networks underlie cell-state heterogeneity in disease progression. Together, these results highlight STAGED as a general framework for uncovering biologically grounded mechanisms of cellular communication and decision-making, bridging mechanistic modeling and deep learning to advance our understanding of multicellular systems. |
| 10:30-11:00pm | Break (30 minutes) |
| 11:00-11:30pm | Didactic: Andriy Marusyk: "Embracing the complexity of the evolution of targeted therapy resistance"Targeted therapies directed against oncogenic signaling addictions, such as ALK inhibitors (ALKi) in ALK+ NSCLC, tend to induce strong and durable clinical responses. However, targeted therapies are not curable, as subsets of tumor cells survive, persisting under therapy. Subsequently, these residual populations evolve resistance, manifested as cancer relapse. Massive research and development efforts have been directed to overcoming resistance and achieving cures. However, despite identifying large numbers of molecular drivers of resistance and extending remission times, targeted therapies remain non-curable, and resistance remains inevitable. The relatively slow pace of therapeutic progress reflects multiple challenges, including both well-appreciated and emerging ones. Correct identification of the challenges is essential for directing research efforts necessary for moving from incremental progress to major breakthroughs. In addition to the obvious challenges of inter-tumor heterogeneity, identification of specific resistance mechanisms and drug availability, the once emergent challenge of intra-tumor heterogeneity is generally well appreciated. Less recognized is the fundamental challenge of evolvability of neoplastic populations, i.e., the ongoing ability to generate heritable phenotypic variability that can fuel the evolution of resistance. Even less appreciated is the emergent multifactorial nature of therapy persistence and resistance, i.e., that, even at the level of individual tumor cells, persistence and resistance are not necessarily reducible to a single mechanistic driver. Instead, the evolution of resistance develops as trajectories in adaptive and epigenetic landscapes, shaped by an integrated input of multiple tumor cell intrinsic, microenvironmental, and systemic factors. Addressing the fundamental issue of therapy resistance requires the acknowledgment of the complex spatiotemporal dynamics of evolving resistance and the development of therapeutic strategies that consider integrative inputs that shape loss of therapy sensitivity while hindering the ability of tumors to adapt. |
| 11:30-11:45am | Natalie Meacham: "An Inverse Problem to Recover Sensitivity to Treatment in Cyclically Treated Prostate Cancer"Resistance to treatment, which comes from the heterogeneity of cell types within tumors, is a leading cause of poor treatment outcomes in cancer patients. Previous mathematical work modeling cancer over time has neither emphasized the relationship between cell heterogeneity and treatment resistance nor depicted heterogeneity with sufficient nuance. To respond to the need to depict a wide range of resistance levels, we develop a random differential equation model of tumor growth. In the inverse problem, we aim to recover the sensitivity to treatment as a probability mass function. This allows us to observe what proportions of cells exist at different sensitivity levels. After validating the method with synthetic data, we apply it to in vitro monoclonal and mixture cell population data of isogenic Ba/F3 murine cell lines to uncover each tumor's levels of sensitivity to treatment as a probability mass function. We then expand the model to synthetic tumors being treated cyclically, successfully recovering the sensitivity distribution from each cycle. |
| 11:45-12:00pm | Jeffrey West: "75 years of Math Onco"The integration of advanced mathematical and computational methodologies is playing an increasingly central role in cancer research, offering critical insights connecting biological mechanisms to clinical applications and predictive models that guide clinical decision-making. This growing interest underpins the emergence of Mathematical Oncology—a multidisciplinary field dedicated to understanding tumor dynamics and optimizing therapeutic strategies through mathematical modeling and computational simulation. The field’s expansion is also reflected in the widespread engagement with platforms such as The Mathematical Oncology website and newsletter, which currently reach over 2,300 subscribers within the scientific community. Yet, the highly interdisciplinary nature of the discipline makes it difficult to have a comprehensive overview of the evolution of the area, and of the different research lines shaping its development. In this study, we address this gap through a comprehensive bibliometric analysis of over 19,000 citable documents across 75 years, and a comparative analysis of hand-curated publications list from 7 years of the Math Onco newsletter. In the analysis, we review the number and type of publications, the impact and topical focus of common journals, common modeling methods over time, the most studied disease sites, and a geographical analysis of institutions involved. Finally, we perform a citation network analysis of the highly productive authors (by total publications) and then illustrate a clustering analysis to group the field into subsections based on disease, modeling methods, and target journals. |
| 12:00-1:00pm | Panel discussion |
| 1:00pm | Meeting Ends |