BayesCOOP is a scalable Bayesian framework for supervised multimodal data integration. It combines the Bayesian bootstrap with a jittered group spike-and-slab Laplace prior to enable cooperative learning across heterogeneous data modalities, with principled uncertainty quantification.