Prototyping a self-learning digital twin platform for personalized treatment in melanoma patients

Executive Summary Cancer patient digital twins (CPDTs)—personalized dynamical simulation models that can make use of high performance computing (HPC) to forecast individual prognosis, assess and optimize treatment options, and benchmark treatment success against a patient's virtual untreated control—could revolutionize cancer care. While advances have been made in high performance scientific computing, machine learning, and multiscalecancer modeling, there has been limited progress to date in bringing CPDTs from theory to clinical reality. This six-month pilot project will bootstrap an aspirational vision to develop clinically-actionable CPDTs for planning immunotherapy in metastatic melanoma patients, with the aim of beginning clinical trials within ten years. The project will leverage cutting-edge multiscale models of heterogeneous tumors, local immune system dynamics, and systems-scale immune expansion to create multiscale model of melanoma metastases in lung tissue and their interactions with the immune system (Milestone 1) and gather canine patient data to help drive rapid model testing (Milestone 2). HPC-driven model exploration will ensure that the multiscale model can recapitulate essential clinical trajectories including spontaneous regression, arrest at sub-clinical size, and growth to clinical detection. Artificial Intelligence (AI)-based techniques will be used to analyze the simulated patient trajectories and develop the first CPDT templates—the first step in fitting a personalized model to an individual patient (Milestone 3). The six-month pilot will develop and test a model of autologous cancer vaccine immunotherapies (Milestone 4) to prepare for prototyping and testing against canine data, while also analyzing patient clinical records to identify high-quality clinical training data and select simulated patient demographics and treatments based upon current clinical needs (Milestone 5). An optional funding level of the pilot would extend the model system to detailed 3D lung tissues in an MPI parallelized framework that is adapted to next-generation HPC platforms (extended Milestones 1e-2e), and perform cutting-edge HPC investigations to evaluate the model performance, identify key model parameters, perform uncertainty quantification, and analyze future data needs (extended Milestones 3e-4e). This project aims to build an open, collaborative community that engages multiple disciplines to realize its vision. All work will be shared as open source and we will encourage the community contributions to help us rapidly advance towards the ideal of CPDTs to improve patient outcomes and quality of life.Executive SummaryCancer patient digital twins (CPDTs)—personalized dynamical simulation models that can make use of high performance computing (HPC) to forecast individual prognosis, assess and optimize treatment options, and benchmark treatment success against a patient's virtual untreated control—could revolutionize cancer care. While advances have been made in high performance scientific computing, machine learning, and multiscalecancer modeling, there has been limited progress to date in bringing CPDTs from theory to clinical reality.This six-month pilot project will bootstrap an aspirational vision to develop clinically-actionable CPDTs for planning immunotherapy in metastatic melanoma patients, with the aim of beginning clinical trials within ten years. The project will leverage cutting-edge multiscale models of heterogeneous tumors, local immune system dynamics, and systems-scale immune expansion to create multiscale model of melanoma metastases in lung tissue and their interactions with the immune system (Milestone 1) and gather canine patient data to help drive rapid model testing (Milestone 2). HPC-driven model exploration will ensure that the multiscale model can recapitulate essential clinical trajectories including spontaneous regression, arrest at sub-clinical size, and growth to clinical detection. Artificial Intelligence (AI)-based techniques will be used to analyze the simulated patient trajectories and develop the first CPDT templates—the first step in fitting a personalized model to an individual patient (Milestone 3). The six-month pilot will develop and test a model of autologous cancer vaccine immunotherapies (Milestone 4) to prepare for prototyping and testing against canine data, while also analyzing patient clinical records to identify high-quality clinical training data and select simulated patient demographics and treatments based upon current clinical needs (Milestone 5).An optional funding level of the pilot would extend the model system to detailed 3D lung tissues in an MPI parallelized framework that is adapted to next-generation HPC platforms (extended Milestones 1e-2e), and perform cutting-edge HPC investigations to evaluate the model performance, identify key model parameters, perform uncertainty quantification, and analyze future data needs (extended Milestones 3e-4e).This project aims to build an open, collaborative community that engages multiple disciplines to realize its vision. All work will be shared as open source and we will encourage the community contributions to help us rapidly advance towards the ideal of CPDTs to improve patient outcomes and quality of life.