Abstract 5025: Applying deep learning transformers and expression quantitative trait loci to interrogate immunogenetic determinants of cancer risk

Abstract The immune system plays a pivotal role in suppressing cancer development through a process known as cancer immunosurveillance. However, substantial immune variation among healthy individuals suggests that the effectiveness of the immune system in eliminating cancer is very likely to vary. This study aims to identify genetic immune characteristics that either protect against or increase cancer risk and to understand how these immune factors interact with known environmental and lifestyle-related risk factors. 
 To achieve this, we have employed a deep learning approach, leveraging transformer models. Transformers can capture complex, non-linear interactions between features, accommodate multimodal input data and incorporate prior knowledge through self-supervised pre-training, which are all relevant properties for our application. As a discovery dataset, we have utilized the UK Biobank, a comprehensive longitudinal cohort consisting of half a million individuals with over 15 years of follow-up. Immune characteristics were derived from plasma proteomics as well as whole genomic sequencing data, in the form of immune cell-specific expression quantitative trait loci (eQTLs). Immune-cell specific eQTLs have been associated with multiple diseases and can provide mechanistic insight into the genetic basis of cancer risk. Here, we share our analysis in exploring immunogenetic factors modulating cancer risk by integrating advanced deep learning techniques with extensive genomic and environmental data. We hope this work will contribute to enhancing our understanding of individual susceptibility to cancer and thereby inform risk stratification strategies and personalized approaches to cancer prevention and intervention, leading to better health outcomes. Citation Format: Sophie Herszterg, Gustavo Arango-Argoty, Etai Jacob. Applying deep learning transformers and expression quantitative trait loci to interrogate immunogenetic determinants of cancer risk [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5025.