A comparative study of deconvolution methods for RNA-seq data under a dynamic testing landscape

Abstract

AbstractDeconvolution analyses have been widely used to track compositional alternations of cell-types in gene expression data. Even though numerous novel methods have been developed in recent years, researchers are still having difficulty selecting optimal deconvolution methods due to the lack of comprehensive benchmarks relative to the newly developed methods. To systematically reveal the pitfalls and challenges of deconvolution analyses, we studied the impact of several technical and biological factors such as simulation model, quantification unit, component number, weight matrix, and unknown content by constructing three benchmarking frameworks that cover comparative analysis of 11 popular deconvolution methods under 1,766 conditions. We hope this study can provide new insights to researchers for future application, standardization, and development of deconvolution tools on RNA-seq data.