Comparison of scRNA-seq data analysis method combinations.

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) generates complex transcriptome data.
• scRNA-seq analysis involves normalization, dimensionality reduction, and clustering.
• Zero counts in scRNA-seq data, often due to dropout events, pose analytical challenges.

Purpose of the Study:

• To evaluate various combinations of established scRNA-seq analysis methods.
• To identify optimal technology combinations for accurate and efficient scRNA-seq data analysis.
• To assess the impact of method combinations on preserving cellular heterogeneity.

Main Methods:

• Summarized two normalization, two dimensionality reduction, and three clustering methods.
• Created 12 distinct technology combinations for scRNA-seq analysis pipelines.
• Evaluated combinations using the Goolam scRNA-seq dataset (ArrayExpress accession E-MTAB-3321).

Main Results:

• Identified specific technology combinations that enhance scRNA-seq analysis efficiency and accuracy.
• Demonstrated that optimized combinations effectively reduce technical noise and compress data.
• Showcased that selected combinations preserve crucial cell heterogeneity for downstream applications.

Conclusions:

• The choice of method combinations significantly impacts scRNA-seq data analysis outcomes.
• Appropriate technology combinations are essential for robust noise reduction, dimensionality reduction, and cell clustering.
• Optimized scRNA-seq analysis pipelines ensure reliable identification of cellular heterogeneity.