Population-scale integration of spatial omics networks for clinical prediction and biological discovery by SPIN.

SPIN integrates spatial single-cell co-expression/co-abundance networks to predict dementia status in the SEA-AD MERFISH cohort with accuracy 0.81.

• Why it matters: Spatial omics links molecular organization to clinical outcomes, but population-scale models that incorporate spatial single-cell information remain difficult to build and interpret.
• What they did: SPIN constructs subject-specific, cell-type level spatial networks and uses Bayesian scalar-on-network regression with manifold learning (BSNMani) for molecular-to-clinical prediction.
• The result: In SEA-AD, SPIN also uncovers biologically relevant co-expression subnetworks, and in a breast cancer single-cell proteomics cohort it predicts survival with C-index 0.78.

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