tooluniverse-multi-omics-integration

Match common samples across both matrices, map features to gene-level identifiers, then compute Spearman correlation per gene. Typical RNA-protein correlations range from 0.4 to 0.6; genes with |r| below 0.2 are flagged as discordant and may indicate post-transcriptional regulation.

MOFA+ identifies latent factors explaining cross-omics variance, joint NMF performs shared decomposition across omics layers, and SNF fuses similarity networks from heterogeneous data types. MOFA+ is best for separating shared versus omics-specific drivers.

Promoter methylation typically represses expression, so expect negative Spearman correlations between CpG beta values and gene expression. Genes with r below -0.5 and p below 0.01 are flagged as epigenetically regulated, though exceptions to the repression model exist.

At least 10 common samples across omics types are required, but 20 or more are recommended for reliable integration. If not all samples have all omics types, use pairwise integration instead of requiring complete intersection.

Batch effects across platforms require careful normalization, large datasets demand significant memory, and results need domain expertise for interpretation. Discordant RNA-protein pairs may reflect real post-transcriptional regulation rather than noise.