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

• Published in: Briefings in bioinformatics Vol. 25; no. 5
• Main Authors: Li, Minghan, Su, Yuqing, Gao, Yanbo, Tian, Weidong
• Format: Journal Article
• Language: English
• Published: England Oxford University Press 25.07.2024; Oxford Publishing Limited (England)
• Subjects: Algorithms; Breast cancer; Computational Biology - methods; Estimation; Female; Humans; Lymphocytes; Lymphocytes T; Medical research; Prognosis; Robustness; Triple Negative Breast Neoplasms - pathology; subject-specific references; rare cell types; deconvolution; RNA-seq
• ISSN: 1467-5463; 1477-4054; 1477-4054
• DOI: 10.1093/bib/bbae422

Abstract In this study, we introduce Robust estimation of Cell type proportions by Integrating single-reference-based DEconvolutions (ReCIDE), an innovative framework for robust estimation of cell type proportions by integrating single-reference-based deconvolutions. ReCIDE outperforms existing approaches in benchmark and real datasets, particularly excelling in estimating rare cell type proportions. Through exploratory analysis on public bulk data of triple-negative breast cancer (TNBC) patients using ReCIDE, we demonstrate a significant correlation between the prognosis of TNBC patients and the proportions of both T cell and perivascular-like cell subtypes. Built upon this discovery, we develop a prognostic assessment model for TNBC patients. Our contribution presents a novel framework for enhancing deconvolution accuracy, showcasing its effectiveness in medical research.