Identifying Cell Subpopulations and Their Genetic Drivers from Single-Cell RNA-Seq Data Using a Biclustering Approach

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-Seq) provides high-resolution cellular activity data.
• Analyzing scRNA-Seq data is crucial for understanding cell heterogeneity in development and disease.
• Identifying cell subpopulations and their genetic drivers is a key challenge in scRNA-Seq analysis.

Purpose of the Study:

• To present BiSNN-Walk, a fast and simple iterative biclustering algorithm for scRNA-Seq data.
• To enhance cell clustering and gene association analysis for improved biological interpretability.
• To introduce an entropy-based measure for selecting optimal similarity matrices.

Main Methods:

• Developed BiSNN-Walk, an iterative biclustering approach building on the SNN-Cliq algorithm.
• Implemented an entropy-based measure for similarity matrix selection.
• Applied BiSNN-Walk to three large-scale scRNA-Seq datasets.

Main Results:

• BiSNN-Walk demonstrated comparable or improved cell clustering performance over SNN-Cliq.
• The algorithm generated biologically meaningful gene clusters, validated by Gene Ontology (GO) term enrichment.
• Top characteristic genes for similar cell states showed significant overlap, confirming gene cluster fidelity.

Conclusions:

• BiSNN-Walk is an effective tool for scRNA-Seq data analysis, offering reliable cell subpopulation identification.
• The algorithm provides interpretable gene rankings, aiding in the discovery of genetic drivers.
• BiSNN-Walk enhances the biological insights obtainable from scRNA-Seq studies.