Batch-effect correction in single-cell RNA sequencing data using JIVE.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Batch effects in single-cell RNA sequencing (scRNA-seq) data, arising from technical variations, obscure true biological signals.
• The Joint and Individual Variation Explained (JIVE) method can disentangle shared biological patterns from batch effects in multi-source data.
• Existing JIVE implementations are computationally intensive and not suitable for large-scale scRNA-seq datasets.

Purpose of the Study:

• To enhance the computational efficiency of JIVE for large-scale scRNA-seq data.
• To develop a novel application of JIVE for batch-effect correction across multiple scRNA-seq datasets.
• To improve the extraction of biological variability while accounting for technical noise.

Main Methods:

• Implemented an computationally efficient version of JIVE tailored for scRNA-seq data.
• Applied the enhanced JIVE to decompose scRNA-seq datasets into joint (biological) and individual (technical) structures.
• Benchmarked the enhanced JIVE against established batch-correction tools: Seurat v5, Harmony, LIGER, and Combat-seq.

Main Results:

• The enhanced JIVE method demonstrated superior performance in preserving cell-type specific effects compared to other tools.
• JIVE excelled in batch-effect correction, particularly in datasets with balanced batch sizes.
• The method successfully separated biological variability from technical variations within batches.

Conclusions:

• The computationally enhanced JIVE is a powerful tool for batch-effect correction in large-scale scRNA-seq data.
• This approach effectively isolates true biological signals, facilitating more accurate downstream analyses.
• The enhanced JIVE offers a robust solution for multi-dataset integration in single-cell genomics.