LMD: Cluster-Independent Multiscale Marker Identification in Single-cell RNA-seq Data.

Area of Science:

• Single-cell RNA sequencing analysis
• Computational biology
• Genomics

Background:

• Accurate cell marker identification is essential for dissecting cellular heterogeneity in single-cell RNA-seq (scRNA-seq) data.
• Existing methods may struggle with fine-grained resolution and cross-sample comparisons without complex integration.

Purpose of the Study:

• To introduce Localized Marker Detector (LMD), a novel computational tool for identifying localized genes.
• To characterize cellular diversity and function in scRNA-seq data with multi-resolution and fine-grained precision.
• To enable robust cross-sample comparisons and identification of novel cell populations.

Main Methods:

• Construction of a cell-cell affinity graph.
• Diffusion of gene expression values across the cell graph.
• Scoring genes based on diffusion dynamics to identify localized markers.
• Grouping candidate markers into functional gene modules.

Main Results:

• LMD successfully identified localized genes that accurately reflect cell types, subtypes, and biological variations like cell cycle status.
• Application to mouse bone marrow and hair follicle dermal condensate datasets revealed shared and sample-specific gene signatures.
• LMD facilitated the discovery of novel cell populations without requiring batch effect correction or integration.
• Performance evaluation across nine scRNA-seq datasets demonstrated LMD's superior performance compared to six other methods.

Conclusions:

• LMD is an effective tool for identifying localized genes and characterizing cellular diversity in scRNA-seq data.
• LMD simplifies cross-sample comparisons and the discovery of novel cell populations.
• LMD offers a robust and high-performing alternative for marker gene identification in scRNA-seq analysis.