Project summary While immune checkpoint inhibitors (ICIs) have transformed cancer therapy, durable response has been observed in only 20-30% of patients with non-small cell lung cancer (NSCLC). The variability in clinical outcomes highlights the urgent and unmet need for novel accurate biomarkers. Human leukocyte antigen (HLA) genes are key to antigen presentation, with their inherited and acquired polymorphisms markedly influencing the diversity of peptide repertoire and individual immune response. Yet, the exact molecular mechanisms giving rise to immunity are not fully defined. The overarching goal of this proposal is to elucidate the role of both germline and somatic HLA variations in cellular immunity and the efficacy of ICI monotherapy in patients with NSCLC, employing novel statistical and machine learning (ML) methods to analyze multi-omic and clinical data. I propose to leverage the unique and extensive multi-omic (i.e., germline genetics, tumor genomics, transcriptomics, and proteomics) and clinical data at the Dana-Farber Cancer Institute, Memorial Sloan Kettering Cancer Center, Stand Up To Cancer (SU2C), The Cancer Genome Atlas (TCGA), and private entities of TEMPUS and CARIS Life Science, including ~40% non-European in over 80K patients with NSCLC, to achieve the following aims. In Aim 1, I will identify the mechanisms among germline and somatic HLA variations (i.e., germline HLA heterozygosity, evolutionary divergence (HED), somatic HLA loss of heterozygosity (LOH), and HLA gene expression), somatic mutations, and cellular immune phenotypes (i.e., PD-L1 expression and immune infiltration) using causal mediation analysis. In Aim 2, I will examine how these germline and somatic HLA variations impact ICI monotherapy through (2a) evaluating the associations with objective response rate (ORR), progression-free survival (PFS), and overall survival (OS) of ICI monotherapy, and (2b) developing and validating a novel weighted neoepitope-based tumor mutational burden (TMB) using patients’ germline HLA alleles and their binding affinities with somatic mutations and evaluate its predictive accuracy of ICI monotherapy. In Aim 3, to accommodate both accuracy and interpretability, I will identify hub genes associated with ORR through weighted gene co-expression network analysis and develop an interpretable multi-omic multimodal ML prediction model for PFS and OS of ICI monotherapy based on selected genes using self-normalizing neural networks. This project's successful execution will identify the mechanisms of germline and somatic HLA variations and immune response in patients with NSCLC to advance precision medicine goals.

Project Narrative While immune checkpoint inhibitors (ICIs) have transformed cancer therapy, durable response has been observed in only 20-30% of patients with non-small cell lung cancer (NSCLC), highlighting the urgent and unmet need for novel accurate biomarkers. Human leukocyte antigen (HLA) genes are key to antigen presentation, with their inherited and acquired polymorphisms markedly influencing the diversity of peptide repertoire and individual immune response. The overarching goal of this proposal is to elucidate the role of germline and somatic HLA variations in cellular immunity and the efficacy of ICI monotherapy in patients with NSCLC, employing novel statistical and machine learning methods to analyze multi-omic and clinical data.