BayesDeBulk: A Flexible Bayesian Algorithm for the Deconvolution of Bulk Tumor Data

Abstract

To understand immune activation and evasion mechanisms in cancer, one crucial step is to characterize the composition of immune and stromal cells in the tumor microenvironment (TME). Deconvolution analysis based on bulk transcriptomic data has been used to estimate cell composition in TME. However, these algorithms are sub-optimal for proteomic data, which has hindered research in the rapidly growing field of proteogenomics. Moreover, with the increasing prevalence of multi-omics studies, there is an opportunity to enhance deconvolution analysis by utilizing paired proteomic and transcriptomic profiles of the same tissue samples. To bridge these gaps, we propose BayesDeBulk, a new method for estimating the immune/stromal cell composition based on bulk proteomic and gene expression data. BayesDeBulk utilizes the information of known cell-type-specific markers without requiring their absolute abundance levels as prior knowledge. We compared BayesDeBulk with existing tools on synthetic and real data examples, demonstrating its superior performance and versatility. AvailabilitySoftware available at http://www.BayesDeBulk.com/ ContactFor any information, please contact francesca.petralia@mssm.edu