RoDPAA: A Robust Dissimilarity-Based Principal Angle Analysis Framework for Multi-View Microbiome Data Integration

RoDPAA aligns microbiome view subspaces using principal angles and Bray–Curtis dissimilarities, recovering shared structure in two IBD cohorts despite sparse, zero-inflated data.

• Why it matters: Integration across body sites or molecular layers is limited by compositional sparsity, modest sample sizes, and methods that only recognize exact sharing.
• What they did: RoDPAA represents each view with any dissimilarity (e.g., Bray–Curtis), aligns signal subspaces via principal angles, and uses permutation testing to quantify shared components.
• The result: In Gaussian simulations it matched JIVE and MOFA with true ranks and recovered the shared dimension in every replicate; in two IBD cohorts it recovered stable shared structure under prevalence filtering, and its leading microbiome–metabolome factors were the only ones associated with IBD.

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