BayICE: A hierarchical Bayesian deconvolution model with stochastic search variable selection

Abstract

16Gene expression deconvolution is a powerful tool for exploring the microenvironment of 17 complex tissues comprised of multiple cell groups using transcriptomic data. Characterizing 18 cell activities for a particular condition has been regarded as a primary mission against diseases. 19For example, cancer immunology aims to clarify the role of the immune system in the 20 progression and development of cancer through analyzing the immune cell components of 21 tumors. To that end, many deconvolution methods have been proposed for inferring cell 22subpopulations within tissues. Nevertheless, two problems limit the practicality of current 23approaches. First, all approaches use external purified data to preselect cell type-specific genes 24 that contribute to deconvolution. However, some types of cells cannot be found in purified 25profiles and the genes specifically over-or under-expressed in them cannot be identified. This 26is particularly a problem in cancer studies. Hence, a preselection strategy that is independent 27 from deconvolution is inappropriate. The second problem is that existing approaches do not 28 recover the expression profiles of unknown cells present in bulk tissues, which results in biased 29 estimation of unknown cell proportions. Furthermore, it causes the shift-invariant property of 30 deconvolution to fail, which then affects the estimation performance. To address these two 31 problems, we propose a novel deconvolution approach, BayICE, which employs hierarchical 32Bayesian modeling with stochastic search variable selection. We develop a comprehensive 33Markov chain Monte Carlo procedure through Gibbs sampling to estimate cell proportions, 34gene expression profiles, and signature genes. Simulation and validation studies illustrate that 35BayICE outperforms existing deconvolution approaches in estimating cell proportions. 36Subsequently, we demonstrate an application of BayICE in the RNA sequencing of patients 37with non-small cell lung cancer. The model is implemented in the R package "BayICE" and 38 the algorithm is available for download. 39 40