Abstract 3165: ImmuneSig: A deep learning framework to develop gene expression signature of immune checkpoint inhibitor therapy

Abstract

Abstract Immune checkpoint inhibitor (ICI) therapy targeting cytotoxic T lymphocyte-associated protein 4 (CTLA-4), programmed death 1 (PD1), and programmed death-ligand 1 (PDL1) induces durable remission in some patients with various cancer types. Given the variable patient outcomes, many studies have been conducted to identify prognostic biomarkers associated with ICI response. However, due to intra- and inter-tumor heterogeneity and the complexity of the immune system, those biomarkers have diverse efficacy for predicting outcomes in clinical trials. Here we propose a deep learning framework with unsupervised self-learning autoencoders that identifies predictive features from patient transcriptomics data. This approach allows us to determine the contribution of each gene to the response classifier and identifies potential gene signatures for predicting immunotherapy response by using the learned weights of trained autoencoders. We applied our framework to 20+ ICI clinical cohorts across different cancer types and compared the prediction performance of our method with other well-known immune response markers evaluated using a suite of machine learning algorithms (random forest, logistic regression, and XGboosting) with cross-validation. We found that in some cohorts, the learned features from autoencoders improve the predictive power of ICI response when evaluated with the area under the operating characteristic curve. Our framework uncovers potential immune gene signatures associated with patient outcomes. Our work will help advance cancer immunology and immuno-oncology research, provide insights into tumor immunity, and lead to the discovery of progressively more immune markers and novel immune targets. Citation Format: Yang Liu, Aashna Jhaveri, Sudhenshna Bodapati, Cheryl Wong, Franziska Michor. ImmuneSig: A deep learning framework to develop gene expression signature of immune checkpoint inhibitor therapy [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 3165.