Predicting Tumor Response to Drug Treatment
AI/Machine Learning/Deep Learning
Scientific Software Development
TimeWednesday, June 19th11am - 11:30am
DescriptionThe DOE-NCI Joint Design of Advanced Computing Solutions for Cancer (JDACS4C) program has set forward three pilot challenge problems. Through the Exascale Computing Project (ECP), The Exascale Deep Learning and Simulation Enabled Precision Medicine for Cancer application development project focuses on the machine learning aspect of the three problems and, in particular, builds a single scalable deep neural network application, CANDLE (CANcer Distributed Learning Environment), that can be used to address each challenge. The National Cancer Institute challenges which CANDLE addresses are (1) understanding the molecular basis of key protein interactions, (2) developing predictive models for drug response, and (3) automating the analysis and extraction of information from millions of cancer patient records to determine optimal cancer treatment strategies.
In this presentation, we discuss a recent computational experiment performed using CANDLE on the Summit supercomputer at the Oak Ridge Leadership Computing Facility to demonstrate how HPC is changing the way we view problems in biomedical science. The aim of the experiment was to improve predictive models of tumor response to drug treatments. The workflow consisted of feature selection, model search, model training, cross-study validation, inference, and uncertainty quantification. Each of these experimental elements represents areas of activity that are common to the scientific workflows of the three JDACS4C pilot challenge problems and more generally in particular deep learning.
Strategic Program Manager