ReCIDE: robust estimation of cell type proportions by integrating single-reference-based deconvolutions

Area of Science:

• Computational Biology
• Bioinformatics
• Cancer Research

Background:

• Accurate estimation of cell type proportions in bulk tissue is crucial for understanding complex biological systems and disease mechanisms.
• Existing deconvolution methods face challenges, especially in accurately quantifying rare cell populations.

Purpose of the Study:

• To introduce ReCIDE (Robust estimation of Cell type proportions by Integrating single-reference-based DEconvolutions), a novel framework to enhance the accuracy of cell type proportion estimation.
• To evaluate ReCIDE's performance against existing methods using benchmark and real-world datasets.
• To apply ReCIDE to triple-negative breast cancer (TNBC) data to identify prognostic biomarkers.

Main Methods:

• Development of the ReCIDE framework, integrating single-reference-based deconvolution approaches.
• Benchmarking ReCIDE against established deconvolution tools on diverse datasets.
• Exploratory analysis of public triple-negative breast cancer (TNBC) bulk RNA-sequencing data using ReCIDE.

Main Results:

• ReCIDE demonstrates superior performance compared to existing methods, particularly in estimating proportions of rare cell types.
• Analysis of TNBC data revealed significant correlations between T cell and perivascular-like cell proportions and patient prognosis.
• A novel prognostic assessment model for TNBC patients was developed based on these findings.

Conclusions:

• ReCIDE offers a robust and accurate framework for cell type deconvolution, advancing the field of computational biology.
• The identified cell type proportions in TNBC are promising prognostic indicators, paving the way for improved patient stratification and treatment strategies.
• This work highlights the utility of advanced deconvolution techniques in uncovering critical biological insights and developing clinical applications.