Benchmarking Algorithms for Gene Set Scoring of Single-cell ATAC-seq Data.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Gene set scoring (GSS) is crucial for analyzing gene expression data (RNA sequencing) to understand cellular heterogeneity.
• Single-cell ATAC sequencing (scATAC-seq) offers insights into gene regulation, but dedicated GSS tools are scarce.
• The applicability of existing RNA sequencing GSS tools to scATAC-seq data requires thorough investigation.

Purpose of the Study:

• To benchmark the performance of various GSS tools on scATAC-seq data.
• To assess the suitability of RNA sequencing GSS tools for scATAC-seq analysis.
• To provide guidelines for selecting appropriate GSS methods and preprocessing techniques for scATAC-seq data.

Main Methods:

• Systematic benchmarking of ten GSS tools (four for bulk RNA-seq, five for scRNA-seq, one for scATAC-seq).
• Evaluation using matched scATAC-seq and scRNA-seq datasets, and up to ten independent scATAC-seq datasets.
• Analysis of the impact of gene activity conversion, dropout imputation, and gene set collections on GSS results.

Main Results:

• GSS tools showed comparable performance on scATAC-seq and scRNA-seq data, indicating their potential applicability.
• Dropout imputation significantly enhanced the performance of most GSS tools for scATAC-seq data.
• The influence of gene activity conversion and gene set choice varied depending on the specific GSS tool and dataset.

Conclusions:

• Existing GSS tools, particularly those for RNA sequencing, can be effectively applied to scATAC-seq data.
• Dropout imputation is a critical preprocessing step for improving GSS performance in scATAC-seq analysis.
• The study provides practical recommendations for optimizing GSS in scATAC-seq studies based on tool and dataset characteristics.