Flexible experimental designs for valid single-cell RNA-sequencing experiments allowing batch effects correction.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for understanding cellular heterogeneity but suffers from technical artifacts like batch effects and dropout events.
• Current methods often rely on idealized experimental designs or complex normalization, limiting their practical application.
• Batch effects obscure true biological variation, while dropout events lead to underestimation of gene expression.

Purpose of the Study:

• To develop a robust computational method for correcting batch effects and handling dropout events in scRNA-seq data.
• To demonstrate the efficacy of the proposed method under realistic experimental designs.
• To provide an interpretable model that integrates multiple scRNA-seq data challenges.

Main Methods:

• Development of Batch effects correction with Unknown Subtypes for scRNA-seq data (BUSseq), a Bayesian hierarchical model.
• Mathematical proof of batch effect separation under reference panel and chain-type designs.
• Simultaneous clustering, batch correction, imputation, and differential gene expression analysis without prior normalization.

Main Results:

• BUSseq effectively separates biological variability from batch effects in scRNA-seq data.
• The method demonstrates superior performance compared to existing approaches on both simulated and real datasets.
• BUSseq successfully clusters cell types, imputes missing values, and identifies differentially expressed genes.

Conclusions:

• BUSseq offers a powerful and flexible solution for common challenges in scRNA-seq data analysis.
• The method enhances the reliability and interpretability of scRNA-seq experiments, even with non-ideal designs.
• BUSseq advances the field by providing an integrated approach to data correction and biological discovery.