bioRxiv2023
Gene expression studies often use bulk RNA sequencing of mixed cell populations because single cell or sorted cell sequencing may be prohibitively expensive. However, mixed cell studies may miss expression patterns that are restricted to specific cell populations. Computational deconvolution can be used to estimate cell fractions from bulk expression data and infer average cell-type expression in a set of samples (eg cases or controls), but imputing sample-level cell-type expression is required for more detailed analyses, such as relating expression to quantitative traits, and is less commonly addressed. Here, we assessed the accuracy of imputing sample-level cell-type expression using a real dataset where mixed peripheral blood mononuclear cells (PBMC) and sorted (CD4, CD8, CD14, CD19) RNA sequencing data were generated from the same subjects (N=158), and pseudobulk datasets synthesised from eQTLgen single cell RNA-seq data. We compared three domain-specific methods, CIBERSORTx, bMIND and debCAM/swCAM, and two cross-domain machine learning methods, multiple response LASSO and ridge, that had not been used for this task before. We also assessed the methods according to their ability to recover differential gene expression (DGE) results. LASSO/ridge showed higher sensitivity but lower specificity for recovering DGE signals seen in observed data compared to deconvolution methods, although LASSO/ridge had higher area under curves than deconvolution methods. Machine learning methods have the potential to outperform domain-specific methods when suitable training data are available. Author SummaryNumerous studies have demonstrated that gene expression in particular subsets of immune cells plays a critical role in the development of diseases and response to treatment. By profiling gene expression from these cells, we can identify disease-relevant genes, comprehend their functions in the disease or response to treatment, and potentially pave the way for screening and patient stratification for prevention and treatment. However, the current cost of single-cell RNA sequencing is too high for large-scale expression profiling analysis. Therefore, an alternative approach is to computationally estimate cell-type specific expression from mixed cell populations, which has been less explored in the field. With this in mind, we proposed using machine learning approaches, multiple response LASSO and ridge, and applied them to synthesised datasets and real-world data where gene expression was measured in mixed and pure cell populations of the same subjects. We compared them to standard methods in the field, and evaluated the accuracy of predicted expression as well as the ability to reconstruct differentially expressed gene signals. Our results revealed that the LASSO/ridge algorithms performed better than existing methods in recovering differentially expressed gene signals, highlighting their potential applications to impute the cell-type expression.