scSelector: A Flexible Single-Cell Data Analysis Assistant for Biomedical Researchers.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Standard single-cell RNA sequencing (scRNA-seq) analysis relies on rigid clustering methods, potentially missing subtle cellular differences.
• Stringent quality control (QC) filtering in scRNA-seq can lead to the loss of biologically significant cells.
• Existing workflows lack flexibility for researchers to interactively define and analyze cell populations based on biological expertise.

Purpose of the Study:

• To develop scSelector, an interactive software toolkit for flexible scRNA-seq data analysis.
• To enable researchers to select and analyze cell populations directly from low-dimensional embeddings using biological knowledge.
• To integrate AI-driven interpretation for enhanced cell population characterization.

Main Methods:

• Developed scSelector using Python with Scanpy, Matplotlib, and NumPy.
• Integrated an interactive lasso selection tool with differential expression and functional enrichment analysis modules.
• Incorporated Large Language Model (LLM) assistance for automated cell-type and state prediction reports.

Main Results:

• scSelector successfully resolved functional heterogeneity within cell types, identifying distinct alpha-cell subpopulations.
• Characterized rare cell populations, including platelets in PBMCs and low-abundance endothelial cells.
• Demonstrated that cells discarded by standard QC can be biologically functional, and identified states of outlier cells like proliferative NK cells.

Conclusions:

• scSelector offers a flexible, researcher-centric alternative to automated scRNA-seq pipelines.
• The combination of interactive selection and AI interpretation improves scRNA-seq analysis precision.
• Facilitates the discovery of novel cell types and complex cellular behaviors.