Abstract
Plasma cell-free RNA (cfRNA) is derived from cells in various tissues and organs throughout the body and reflects the physiological and pathological conditions. Identifying the origins of cfRNA is essential for comprehending its variations. Only a few tools are designed for cfRNA deconvolution, and most studies have relied on traditional bulk RNA methods. In this study, we employed human tissue and cell transcriptomic data as reference sets and evaluated the performance of seven deconvolution methods on cfRNA. We compared the analysis results of cell types and tissues of origin of plasma cfRNA and chose to use single-cell RNA sequencing (scRNA-seq) data as reference to conduct further evaluation of deconvolution methods. Subsequently, we assessed the accuracy and robustness of the methods by utilizing simulated cfRNA data generated from scRNA-seq. We also evaluated the methods' accuracy on real plasma cfRNA data by analyzing the correlation between the predicted cell proportions and the corresponding clinical indicators. Moreover, we compared the methods' effectiveness in revealing the impacts of diseases on cells and evaluated the performance of cancer classification models based on the cell origin data they provided. In summary, our study provides valuable insights into cfRNA origin analysis, enhancing its potential in biomedical research.