MS05 - ONCO-03

MathOnco Subgroup Mini-Symposium: At the Interface of Modeling and Machine Learning (Part 1)

Wednesday, July 16 at 10:20am

SMB2025 Follow

Note: this minisymposia has multiple sessions. The other session is: and Part-2.

Jana Gevertz (The College of New Jersey), Thomas Hillen (University of Alberta), Linh Huynh (Dartmouth College)

Description:

Mathematical oncology models describe cancer dynamics using biologically motivated equations that are validated using experimental data. Machine learning models, on the other hand, leverage vast amounts of data to make predictions without necessarily including any a-priori biological knowledge. Mathematical models result in biologically interpretable predictions, whereas machine learning models excel at handling complex, high-dimensional datasets. Thus, work at the interface of modeling and machine learning holds the promise of realizing the advantages of both methods. In this mini-symposium, we will showcase how cancer research benefits from a combined approach of mathematical modeling and machine learning.

Thomas E. Yankeelov

The University of Texas at Austin

"Integrating mechanism-based and data driven modeling to predict breast cancer response to neoadjuvant chemotherapy"

While neoadjuvant chemotherapy (NAC) is the standard-of-care for treatment of patients with triple-negative breast cancer (TNBC), only about half of patients attain a pathological complete response (pCR). The addition of immunotherapy increases the pCR rate to about two thirds, but also increases toxicity. Thus, to improve patient outcomes, it is essential to develop methods that can predict and optimize patient response early during NAC. Previously, we showed that a biology-based model calibrated to pre- and on-treatment patient specific MRI data from the ARTEMIS trial (NCT02276443) can accurately predict patient response to NAC. This model describes cell movement and drug induced death globally over the tumor and cell proliferation locally on a voxel-by-voxel basis. We explore two approaches to extend this model. First, we train a convolutional neural network to predict the calibrated model parameters using only pre-treatment MRI data as input. Second, we apply k-means clustering to the longitudinal MRI data to segment tumors into distinct regions called habitats. Then, rather than calibrating cell proliferation locally, we calibrate one proliferation per habitat. To evaluate these models, we use the total tumor cellularity after the first course of NAC in a receiver operating characteristic (ROC) curve analysis to predict pCR status at the end of NAC. For the first method, we obtain an area under the ROC curve (AUC) of 0.72 for predicting the eventual response to NAC before initiating NAC. For the second method, we obtain AUC values of 0.79 and 0.77 for the habitat-informed and voxel-by-voxel proliferation rate calibrations, respectively, for 101 patients. Thus, we can make reasonably accurate predictions of pCR before initiating NAC with our CNN approach and attain comparable accuracy to a local calibration using our habitats approach, which requires fewer model parameters.

Lena Podina

University of Waterloo

"Universal Physics-Informed Neural Networks and Their Applications"

Differential equations are widely used to model systems such as predator-prey interactions, and the effect of chemotherapy on cancer cells. However, in order to construct these models, assumptions must be made about the behaviour of these systems, which may require significant manual distillation of the literature if the model is large. In this talk, I will discuss Universal Physics-Informed Neural Networks (UPINNs), and show how UPINNs can be used to learn unknown terms in ordinary and partial differential equations from sparse and noisy data. This approach allows one to use machine learning to identify the best way to model a system, rather than relying on prior assumptions. Physics Informed Neural Networks (PINNs) have been very successful in a sparse data regime (reconstructing entire ODE solutions using only a single point or entire PDE solutions with very few measurements of the initial condition). The Universal PINN approach (UPINN) adds a neural network that learns a representation of unknown hidden terms in the differential equation. These hidden term neural networks can then be converted into symbolic equations using symbolic regression techniques like AI Feynman. In our work, we demonstrate strong performance of UPINNs even when provided with very few measurements of noisy data in both the ODE and PDE regime. We apply UPINNs to learning predator-prey interaction in the Lotka-Volterra model, chemotherapy drug action terms in a model of cancer cell growth, and terms in Burgers’ PDE. UPINNs could be instrumental to paving the way to allow machine learning to help applied mathematicians model systems in a more automatic, data-driven way even when observations are sparse.

Adam L. MacLean

University of Southern California

"Dynamic rewiring of cell-cell interaction networks in metastatic TMEs to empower checkpoint inhibition"

Tumors grow, evolve, and metastasize as a result of an intricate set of interactions between the numerous cell types that comprise the tumor microenvironment (TME). Many of these interactions remain poorly understood, even in the absence of therapy, and the size of the networks that must be considered necessitates a systems biology approach. Immune checkpoint inhibition combined with entinostat, a histone deacetylase inhibitor, has been shown to promote durable responses in a minority of patients with metastatic, triple negative breast cancer. But the mechanisms of action of entinostat at metastatic sites has not been investigated. We measured the molecular properties of the metastatic TME in high resolution via scRNA-seq, quantifying 39 cell states across six treatment arms. Entinostat treatment led to an increase in stemness in tumor cells and decreased mesenchymal gene expression. To study the wiring of cell-cell interaction networks with and without treatments we inferred small cell circuits that are over-represented in the full networks. Top ranked cell circuits comprised myeloid and T cell subtypes: interactions between which were dramatically affected by combination therapy. Top pathways contributing to these interactions included chemokine, galectin, and ICAM pathways, which we tested via targeting specific ligand-receptor pairs in functional suppression assays, coupled to predictions from mathematical models to simulate therapeutic responses for a given treatment intervention. Beyond the clinical impact of this work, our results offer a framework with which to decompose large, complex TMEs to infer multiscale networks mediating treatment effects and then to infer the tumor response dynamics via mathematical modeling.

Venkata Manem

Centre de Recherche du CHU de Québec; Université Laval, Canada Université Laval, Canada

"Beyond the One-Size-Fits-All Paradigm: Leveraging Bioinformatics and AI to Advance Biomarker-Guided Oncology."

In the era of precision oncology, the convergence of high-throughput technologies and artificial intelligence (AI) is transforming how we understand and treat cancer. Traditional 'one-size-fits-all' regimens are being replaced by biomarker-driven strategies that tailor therapies to individual patients. However, the complexity of tumor genomics and microenvironments limits the effectiveness of many biomarkers, underscoring the need for biologically informed approaches. This talk will explore the intersection of bioinformatics and AI in uncovering biological insights for personalized cancer care. The first part will focus on breast cancer, emphasizing the critical role of epithelium-stroma crosstalk utilizing transcriptomics data to uncover the dynamic interactions within the tumor microenvironment. The second part will address the discovery of biomarkers in non-small cell lung cancer patients treated with immunotherapy, leveraging medical imaging data. Together, these efforts demonstrate how AI and bioinformatics bridge the gap between molecular complexity and actionable insights, paving the way for patient-specific decision-making in oncology.

#SMB2025 Follow

Annual Meeting for the Society for Mathematical Biology, 2025.