Omnibus and robust deconvolution scheme for bulk RNA sequencing data integrating multiple single-cell reference sets...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Cell-type deconvolution of bulk tissue RNA sequencing (RNA-seq) data is crucial for understanding disease-associated variations in cellular composition.
• Existing deconvolution methods often depend heavily on the quality of external data, such as single-cell RNA sequencing (scRNA-seq) references and prior biological knowledge.

Purpose of the Study:

• To present the Integrated and Robust Deconvolution (InteRD) algorithm for inferring cell-type proportions from bulk RNA-seq data.
• To develop a deconvolution method that is robust to inaccuracies in external information and effectively integrates multiple data sources.

Main Methods:

• The InteRD algorithm utilizes penalized regression with a novel evaluation criterion for deconvolution.
• It integrates deconvolution results from multiple scRNA-seq datasets.
• InteRD calibrates reference-based estimates using additional prior biological information.

Main Results:

• InteRD demonstrates enhanced accuracy and robustness in estimating cell-type proportions compared to existing methods.
• Numerical evaluations and real-data applications show that InteRD's estimates align well with biological knowledge.
• The algorithm effectively handles inaccurate external information, improving deconvolution system reliability.

Conclusions:

• The InteRD algorithm provides a more accurate and robust approach to cell-type deconvolution from bulk RNA-seq data.
• Its ability to integrate multiple data sources and incorporate prior biological information enhances biological interpretation.
• InteRD offers a valuable tool for researchers studying cellular heterogeneity in various biological contexts.