spOT-NMF: gene expression deconvolution via optimal transport with proximity-aware niche mapping for spatial transcriptomics.

spOT-NMF uses optimal transport-based non-negative matrix factorization to rank first among reference-free deconvolution methods across five gold-standard spatial transcriptomics datasets, especially under xenograft signal imbalance.

• Why it matters: Spatial transcriptomics needs accurate reference-free deconvolution of mixed signals, including species-admixed xenografts, and tools to map per-spot program usage into spatial niches.
• What they did: The method replaces standard reconstruction loss with Wasserstein distance for OT-NMF factorization, then adds a proximity-based niche inference module to build directional co-occurrence networks from spatial program co-usage.
• The result: It recovers cell populations absent from matched single-cell atlases and, in glioblastoma xenografts and human colorectal cancer, identifies spatially restricted interferon signaling, invasion routes, and CAF–tumor boundaries.

AI summary of the abstract · may contain errors · not medical advice · read the paper.