CATD: A reproducible pipeline for selecting cell-type deconvolution methods across tissues

Abstract

ABSTRACTCell-type deconvolution methods aim to infer cell composition from bulk transcriptomic data. The proliferation of developed methods, coupled with the inconsistent results obtained in many cases, highlights the pressing need for guidance in the selection of appropriate methods. The growing accessibility of systematic single-cell RNA sequencing datasets, often accompanied by bulk expression from related samples, makes it possible to benchmark the existing methods more objectively. Here, we propose a comprehensive assessment of 29 deconvolution methods, leveraging single-cell RNA-sequencing data from different tissues. We evaluate deconvolution across a wide range of simulation scenarios and we show that single-cell regression-based deconvolution methods perform well while their performance is highly dependent on the reference.We also study the impact of bulk-reference differences, including those associated with sample, study, and technology. We provide validation using a gold standard dataset from mononuclear cells and suggest a consensus prediction of proportions when ground truth is not available. We f validated the consensus method on data from the stomach and studied its spillover effect. Lastly, we suggest that the Critical Assessment of Transcriptomic Deconvolution (CATD) pipeline can be employed for simultaneous deconvolution of hundreds of bulk samples and we envision it to be used for speeding up the evaluation of newly developed methods.Key Points– Thorough assessment of 29 deconvolution methods, leveraging diverse single-cell RNA-sequencing data from various tissues, alongside extensive simulations and validation against known ground truth data.– Emphasis on the pivotal role of reference selection, tissue type, and technological nuances in determining the efficacy of deconvolution methods.– Introduction of the user-friendly and robust Critical Assessment of Transcriptomic Deconvolution (CATD) Snakemake pipeline, enabling efficient and reproducible cell-type deconvolution in real bulk RNA-Seq datasets.