Integration-free and cluster-free detection of localized differential gene coexpression in single-cell data with DiCoLo.

DiCoLo detects differential gene colocalization in single-cell data without clustering or cross-condition alignment, robustly identifying localized coordinated expression changes even under weak signals and complex batch effects.

• Why it matters: It addresses the challenge of finding where gene coordination patterns change between conditions, especially when effects are confined to small subpopulations or obscured by batch effects.
• What they did: For each condition, DiCoLo builds a gene graph using Optimal Transport distances over localized cell neighborhoods, then finds differential gene programs by detecting connectivity changes between gene graphs.
• The result: Across multiple benchmark datasets, DiCoLo outperforms existing methods, and in mouse hair follicle development it reveals coordinated gene programs and emerging cell populations linked to perturbations in morphogen signaling.