Controlling for Confounding Effects in Single Cell RNA Sequencing Studies Using both Control and Target Genes

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNAseq) offers high-resolution transcriptome analysis of cell populations.
• Confounding effects can compromise the accuracy of scRNAseq data analysis.
• Robust methods are needed to control for these confounding factors.

Purpose of the Study:

• To introduce scPLS (single-cell partial least squares), a novel statistical method for inferring confounding effects in scRNAseq data.
• To provide a robust and accurate approach for controlling technical and biological confounders.

Main Methods:

• scPLS models two sets of genes: a control set (free of predictor effects) and a target set (of primary interest).
• Utilizes partial least squares regression to jointly model these gene sets, maximizing data utilization for confounder inference.
• Validated through extensive simulations and comparisons with existing methods.

Main Results:

• scPLS demonstrates effectiveness in robustly inferring confounding effects.
• Simulations confirm the superiority of scPLS compared to other methods.
• Application to real scRNAseq datasets shows successful removal of technical and cell cycle effects.

Conclusions:

• scPLS is a powerful tool for improving the accuracy of scRNAseq data analysis.
• The method enhances downstream analyses by effectively mitigating confounding influences.
• scPLS offers significant benefits for researchers studying heterogeneous cell populations.