Accurate and well-powered case-control analysis of spatial molecular data.

VIMA identifies disease-associated spatial features with greater flexibility and precision using an ensemble of variational autoencoders in spatial molecular data.

• Why it matters: Current methods often rely on restrictive assumptions that may overlook important spatial signals in tissue samples, limiting understanding of disease mechanisms.
• What they did: The authors developed VIMA, which trains multiple variational autoencoders to generate numeric tissue 'fingerprints' and defines overlapping microniches for meta-analysis across datasets.
• The result: VIMA demonstrated accurate calibration, power, and spatial precision in simulations and successfully recapitulated known biology while uncovering new disease-related spatial features across three diseases and modalities.