Analysis scripts for: Robust within-participant inference for paired longitudinal mucosa-stool multi-omics (HMP2 ibdmdb cohort).

Paired mucosal biopsy - stool metagenome analysis of the HMP2 ibdmdb cohort: 235 matched biopsy-stool pairs from 84 participants (CD 114 / UC 74 / nonIBD 47 biopsy events; pairing rule = earliest stool sample within +/-14 days of biopsy). Three hypotheses were tested: H1 (microbial, PASS): continuous dysbiosis score (PC1 of 10 disease-associated genera) drives genus-level GEE signatures; Faecalibacterium shows opposite direction in CD vs UC (disease-stratified main finding). H2 (host, PASS): per-gene CR1-clustered SE over 31,126 genes; 2,856 IBD FDR<0.05; 9/9 curated genes up-regulated (CXCL1 beta=0.91, SOCS3, PTGS2, IL17A, ...). H3 (coupling, negative): gene-level 33 hits (64% non-coding, incl. PRSS contamination markers); module-level inflammatory cluster does not split (0.80-0.99 range) -> suggestive, not significant.