Locat: Joint enrichment and depletion testing identifies localized marker genes in single-cell transcriptomics

Area of Science:

• Single-cell transcriptomics
• Computational biology
• Genomics

Background:

• Existing methods for marker gene identification in single-cell data often focus on enrichment within populations.
• These methods may not adequately test for expression depletion outside designated cell clusters.
• This limitation can lead to less specific marker genes and reduced interpretability.

Purpose of the Study:

• To introduce Locat, a novel framework for identifying highly specific localized genes in single-cell transcriptomic data.
• To develop a method that tests for both expression concentration within cell populations and depletion outside of them.
• To enable more precise marker gene discovery for cell population delineation and biological interpretation.

Main Methods:

• Locat fits weighted Gaussian mixture models to gene expression and background densities.
• It computes test statistics for expression concentration in compact regions and depletion elsewhere.
• A unified localization score integrates these statistics for gene ranking.

Main Results:

• Locat successfully identifies localized genes in synthetic data with controlled ground truth, including various expression patterns.
• In biological datasets, Locat discovers compact marker sets capturing lineage, condition-specific, and temporal dynamics.
• Embeddings from localized genes preserve cell population separation and biological trajectories, outperforming traditional methods like highly variable genes selection.

Conclusions:

• Jointly assessing expression concentration and depletion provides specific, interpretable marker genes.
• Locat enables direct cross-condition and multi-sample comparisons of marker genes.
• The framework is effective across diverse biological settings, including developmental, perturbation, and differentiation studies.