Presentation Information
[SS24-01]Deep Learning Model for Inferring Parameters and Hidden Dynamics in Cancer Models
*Rathtish Kumar BV1, Bishal Chhetri1 (1. IIT Kanpur (India))
Keywords:
Cancer Model,Deep Learning,Parameter Inference,PINNS,ODE
Cancer progression and growth are influenced by various interactions within the tumor microenvironment. The concept of cancer hallmarks has guided our understanding of the multiple pathways that drive cancer growth and progression, from cancer cells acquiring self-sufficient growth signals to their ability to evade the immune system. Each pathway of cancer progression involves dynamic changes in the molecular interactions and regulation within the biochemical networks of the tumor microenvironment. To understand the dynamics of cancer growth and progression, these interactions within the tumor micro-environment are usually modelled using ordinary differential equations (ODEs). These system of equations introduce various parameters that need to be estimated accurately and efficiently from the limited noisy experimental measurements. One of the major challenges in modelling these types of biological systems is the accurate estimation of model parameters and the prediction of model dynamics. Among various parameter estimation techniques such as least-squares fitting, genetic algorithms, Bayesian methods, etc., deep learning-based systems-informed neural networks have been introduced and found to be very effective, in inferring the hidden dynamics of experimentally unobserved species and estimation of unknown parameters of the model. In this talk, we will present a physics-informed neural networks approach to infer the dynamics of experimentally unobserved components of a cancer model and estimate the unknown parameters of the model. By incorporating the system of ODEs into the neural networks, we effectively add constraints to the optimization algorithm, which makes the method robust to noisy and sparse measurements.