A0810
Title: BLEND: Bayesian cellular deconvolution with reference selection
Authors: Jiebiao Wang - University of Pittsburgh (United States) [presenting]
Abstract: Cellular deconvolution aims to estimate cell type fractions from bulk transcriptomic and other omic data. While many estimators have been proposed, most of them fail to account for the heterogeneity in cell type-specific (CTS) expression across bulk samples, ignore discrepancies between CTS expression in bulk data and cell type references, and provide no guidance on cell type reference selection or integration. To address these issues, BLEND is introduced, a hierarchical Bayesian method that deconvolves bulk RNA data by leveraging multiple reference datasets. BLEND uses the data to learn the most suitable references for each sample by exploring the convex hulls of references and employs a bag-of-words representation for bulk count data for deconvolution. A Gibbs sampler is derived for posterior computation and an algorithm that maximizes the posterior distribution to speed up computation. The benchmarking studies on both simulated and real human brain data highlight BLEND's superior performance in a variety of scenarios. Requiring no data transformation, cell type marker gene selection, or reference quality evaluation, BLEND facilitates cellular deconvolution with its superior accuracy and robustness.
