Abstract 1160: A machine learning based method for in-vivo metabolic flux analysis of patient tumors

Metabolic fluxes likely control cancer phenotypes and treatment responses but cannot be measured in cancers in human patients. Machine learning has become a popular tool to study cancer biology and can identify complex patterns in the data which may not be decipherable by traditional analyses. We propose an approach to combine 13C-tracer analysis with machine learning to quantify the metabolism of purines in glioblastoma patients. In recent years, multiple studies have been conducted on isotope tracing in cancer patients. While the data for the enrichment of metabolites in human tumors has improved our understanding of cancer metabolism, we lack methods to estimate metabolic fluxes from these enrichment values. A significant challenge in estimating the fluxes in human subjects is that we are limited to enrichment data from a single in-vivo measurement at isotopic non-steady state. We overcome this challenge by combining INST-MFA (Isotopic Non-Stationary-Metabolic Flux Analysis) with machine learning methods. We use the enrichment of circulating metabolites as an input and simulate the time-dependent metabolite enrichment profiles for randomized flux vectors. We use this data to train a convolutional neural network (CNN) model to predict the flux ratios. Apart from being able to predict the fluxes from a single time-point, a machine learning model drastically reduces the time required to fit a traditional INST-MFA model. We demonstrate the real-world validity of the model by predicting flux ratios in mice models and comparing them to results from traditional INST-MFA models. We apply the model to decipher the heterogeneity of the purine pathway in eight patients with brain tumors. Citation Format: Anjali Mittal, Andrew Scott, Abhinav Achreja, Vijay Tarnal, Wajd Al-Holou, Costas Lyssiotis, Daniel Wahl, Deepak Nagrath. A machine learning based method for in-vivo metabolic flux analysis of patient tumors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 1160.