DeSCENT: Deconvolutional Single-Cell RNA-seq Enhances Transcriptome-based Cancer Survival Analysis.

DeSCENT improves survival prediction in eight TCGA cancer cohorts by generating reference-informed cellular profiles from bulk RNA-seq, yielding higher average C-index than all three baselines.

• Why it matters: Tumor heterogeneity from single-cell RNA-seq is seldom available alongside survival outcomes, while bulk cohorts lack direct cellular expression measurements.
• What they did: The method deconvolves bulk RNA-seq into cellular profiles using reference information, then combines bulk and cellular representations to train three survival models.
• The result: Across internal evaluation, DeSCENT generally increased time-dependent AUC and improved average C-index versus Bulk, with mixed integrated Brier score; external validation showed higher average C-index than baselines but lower than internal results.

AI summary of the abstract · may contain errors · not medical advice · read the paper.