Evaluating deconvolution methods using real bulk RNA-expression data for robust prognostic insights across cancer...

Area of Science:

• Computational Biology
• Cancer Research
• Genomics

Background:

• Bulk RNA-expression data analysis is crucial for understanding cancer complexity.
• Traditional pseudobulk benchmarks may lack reliability when absolute cell proportions are unknown.

Purpose of the Study:

• To develop and apply a novel real-data framework for evaluating cancer cell deconvolution methods.
• To identify robust deconvolution algorithms and prognostic cell types in pan-cancer analysis.

Main Methods:

• Leveraged 18 real bulk RNA-expression cohorts (5,891 samples) across nine cancer types.
• Evaluated five deconvolution methods based on differentially proportioned (DP) and prognosis-related (PR) cell types.
• Utilized three benchmark scenarios: consistency with single-cell RNA sequencing (scRNA-seq), cross-cohort reproducibility, and prognostic relevance reproducibility.

Main Results:

• ReCIDE and BayesPrism demonstrated robust performance across benchmark scenarios.
• Identified matrix cancer-associated fibroblasts (mCAF) as a prognostic marker with consistent effects across multiple cancers.
• A combined prognostic indicator of classical monocytes and mCAF proportions was significant in multiple cohorts.

Conclusions:

• Broadened deconvolution benchmarking with a real-data framework.
• Provided actionable tools for precision oncology and guided method selection for translational research.