Decoding cell-type contributions to the cfRNA transcriptomic landscape of liver cancer

Abstract

Background Liquid biopsy, particularly cell-free RNA (cfRNA), has emerged as a promising non-invasive diagnostic tool for various diseases, including cancer, due to its accessibility and the wealth of information it provides. A key area of interest is the composition and cellular origin of cfRNA in the blood and the alterations in the cfRNA transcriptomic landscape during carcinogenesis. Investigating these changes can offer insights into the manifestations of tissue alterations in the blood, potentially leading to more effective diagnostic strategies. However, the consistency of these findings across different studies and their clinical utility remains to be fully elucidated, highlighting the need for further research in this area. Results In this study, we analyzed over 350 blood samples from four distinct studies, investigating the cell type contributions to the cfRNA transcriptomic landscape in liver cancer. We found that an increase in hepatocyte proportions in the blood is a consistent feature across most studies and can be effectively utilized for classifying cancer and healthy samples. Moreover, our analysis revealed that in addition to hepatocytes, liver endothelial cell signatures are also prominent in the observed changes. By comparing the classification performance of cellular proportions to established markers, we demonstrated that cellular proportions could distinguish cancer from healthy samples as effectively as existing markers and can even enhance classification when used in combination with these markers. Conclusions Our comprehensive analysis of liver cell-type composition changes in blood revealed robust effects that help classify cancer from healthy samples. This is especially noteworthy, considering the heterogeneous nature of datasets and the etiological distinctions of samples. Furthermore, the observed differences in results across studies underscore the importance of integrative and comparative approaches in the future research to determine the consistency and robustness of findings. This study contributes to the understanding of cfRNA composition in liver cancer and highlights the potential of cellular deconvolution in liquid biopsy.

Pathway-enhanced Transformer-based robust model for quantifying cell types of origin of cell-free transcriptome

“…cfRNA shows high specificity in early cancer screening, presenting promising prospects [5,21]. Differentiating transcripts of lesion-specific cells from highbackground transcripts originating from blood cells is crucial for early cancer screening and detection.…”

“…2A). Given the non-negligible impact of dropout in cfRNA sequencing [24], we further tested the robustness of the method by evaluating the inference performance of the three methods under different dropouts. This was achieved by randomly dropping expression in these 1,000 simulated cfRNA profiles at rates of 0.1, 0.2, 0.3, 0.4, and 0.5.…”