RIGATonI: An R software for Rapid Identification of Genomic Alterations in Tumors affecting lymphocyte Infiltration

Raven Vella,Emily Hoskins,Lianbo Yu,Julie W Reeser,Michele R Wing,Eric Samorodnitsky,Leah Stein,Elizabeth G Bruening,Anoosha Paruchuri,Michelle Churchman,Nancy Single,Wei Chen,Aharon G Freud,Sameek Roychowdhury
DOI: https://doi.org/10.1101/2024.03.02.583103
2024-06-07
Abstract:Tumor genomic alterations have been associated with altered tumor immune microenvironments and therapeutic outcomes. These studies raise a critical question: are there additional genomic variations altering the immune microenvironment in tumors that can provide insight into mechanisms of immune evasion? This question is the backbone of precision immuno-oncology. Current computational approaches to estimate immunity in bulk RNA sequencing (RNAseq) from tumors include gene set enrichment analysis and cellular deconvolution, but these techniques do not consider the spatial organization of lymphocytes or connect immune phenotypes with gene activity. Our new software package, Rapid Identification of Genomic Alterations in Tumors affecting lymphocyte Infiltration (RIGATonI), addresses these two gaps in separate modules: the Immunity Module and the Function Module. Using pathologist-reviewed histology slides and paired bulk RNAseq expression data, we trained a machine learning algorithm to detect high, medium, and low levels of immune infiltration (Immunity Module). We validated this technique using a subset of pathologist-reviewed slides not included in the training data, multiplex immunohistochemistry, flow cytometry, and digital staining of The Cancer Genome Atlas (TCGA). In addition to immune infiltrate classification, RIGATonI leverages another novel machine learning algorithm for the prediction of gain- and loss-of-function genomic alterations (Function Module). We validated this approach using clinically relevant and function-impacting genomic alterations from the OncoKB database. Combining these two modules, we analyzed all genomic alterations present in solid tumors in TCGA for their resulting protein function and immune phenotype. We visualized these results on a publicly available website. To illustrate RIGATonI's potential to identify novel genomic variants with associated altered immune phenotypes, we describe increased anti-tumor immunity in renal cell carcinoma tumors harboring 14q deletions and confirmed these results with previously published single-cell RNA sequencing. Thus, we present our R package and online database, RIGATonI: an innovative software for precision immuno-oncology research. RIGATonI's software and database can be used for a variety of tasks in precision immuno-oncology including biomarker discovery, drug target identification, and deepening our understanding of tumor-driven mechanisms of immunity.
Bioinformatics
What problem does this paper attempt to address?