PatHoST: weakly supervised reconstruction of pathogen spatial distributions from host transcriptomes

PatHoST reconstructs pathogen spatial distributions from host transcriptomes, and application to human SARS-CoV-2 lung spatial transcriptomics supports generalization across host species and pathogen classes.

• Why it matters: Dual host-pathogen spatial transcriptomics can map infection, but how completely pathogens are detected is unclear, limiting reliable spatial inference.
• What they did: The team benchmarked pathogen detection using pathogen-only spatial transcriptomic sections from a chronic Mycobacterium abscessus lung infection model, then built PatHoST to learn host programs from pathogen-derived silver-standard labels.
• The result: PatHoST reconstructed pathogen spatial patterns across independent tissue sections and identified localized host regulatory programs beyond conventional cluster-based analyses, with human SARS-CoV-2 lung data showing cross-species and cross-pathogen-class generalization.

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