Interpretable single-cell factor decomposition using sciRED.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution gene expression data.
• Interpreting scRNA-seq data is challenging due to technical noise, sparsity, and high dimensionality.
• Existing data factorization methods require manual interpretation of biological signals.

Purpose of the Study:

• To develop an interpretable factor analysis method for scRNA-seq data.
• To improve the identification and characterization of biological signals in scRNA-seq datasets.
• To address limitations in current scRNA-seq data analysis techniques.

Main Methods:

• Developed Single-Cell Interpretable REsidual Decomposition (sciRED).
• sciRED incorporates confounding effect removal, factor rotation for interpretability, and mapping to covariates.
• Identifies unexplained factors and determines associated genes and biological processes.

Main Results:

• Applied sciRED to diverse scRNA-seq datasets (kidney, PBMC, liver).
• Identified sex-specific variation, immune stimulation signals, and reduced ambient RNA contamination.
• Revealed rare cell types and anatomical zonation gene programs.

Conclusions:

• sciRED effectively improves the interpretability of scRNA-seq factor analysis.
• The method aids in characterizing diverse biological signals, including hidden phenomena.
• sciRED offers a valuable tool for advancing scRNA-seq data interpretation and discovery.