Bayesian cell-type deconvolution and gene expression inference reveals tumor-microenvironment interactions

Abstract

Understanding the interactions between cells in their environment is a major challenge in genomics. Here we developed BayesPrism, a Bayesian method to jointly predict cellular composition and gene expression in each cell type, including heterogeneous malignant cells, from bulk RNA-seq using scRNA-seq as prior information. We conducted an integrative analysis of 1,412 bulk RNA-seq samples in primary glioblastoma, head and neck squamous cell carcinoma, and melanoma using single-cell datasets of 85 patients. We identified cell types correlated with clinical outcomes and explored spatial heterogeneity in malignant cell states and non-malignant cell type composition. We refined subtypes using gene expression in malignant cells, after excluding confounding non-malignant cell types. Finally, we identified genes whose expression in malignant cells correlated with infiltration of macrophages, T cells, fibroblasts, and endothelial cells across multiple tumor types. Our work introduces a new lens that uses scRNA-seq to accurately infer cellular composition and expression in large cohorts of bulk data.