Benchmarking differential abundance methods for finding condition-specific prototypical cells in multi-sample...

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Single-cell data analysis uses statistical testing to find gene/protein changes, but noise and heterogeneity cause false positives.
• Differential abundance (DA) testing identifies cell subgroups with significant abundance changes, crucial for understanding disease and experimental impacts.

Approach:

• This study presents the first comprehensive benchmarking of six state-of-the-art DA testing methods.
• Methods were evaluated on synthetic and real single-cell datasets for accuracy, batch effect handling, efficiency, and robustness.

Key Points:

• Accurate identification of true differential abundance subpopulations.
• Effective handling of technical batch effects in single-cell data.
• Assessment of runtime efficiency and hyperparameter usability across methods.

Conclusions:

• Provides dataset-specific recommendations for choosing appropriate DA testing methods.
• Aims to improve the reliability and interpretability of single-cell data analysis.
• Facilitates more robust identification of cell states and biological insights.