Multi-omics Mendelian randomization integrating metabolism, microbiome and immunity supports a putative gut-immune-pelvic pathway in deep infiltrating endometriosis.

Background: Deep infiltrating endometriosis (DIE) is a highly fibrotic and deeply invasive subtype of endometriosis that causes severe pelvic pain, infertility and marked impairment of quality of life. Metabolic, microbial and immune disturbances have been reported in women with endometriosis, but whether these systemic perturbations causally contribute to DIE and which lesion-level molecular mediators connect them to pelvic pathology remains unknown. Methods: We performed two-sample Mendelian randomization (MR) to assess the causal effects of circulating metabolites, gut microbiota (GM) traits and immune cell phenotypes on DIE risk using genome-wide association data from FinnGen and large exposure GWAS. Bayesian colocalization was applied to identify protein-coding genes with shared causal variants between exposures and DIE. Colocalized genes were integrated with RNA-sequencing data from GSE141549 (normal endometrium, n = 43; DIE lesions, n = 88) to evaluate differential expression and immune-cell associations inferred by CIBERSORT-like deconvolution. Machine-learning-based feature selection was used to derive a multigene logistic model, and protein expression of feature genes was validated by immunohistochemistry in independent specimens. Results: MR revealed putative causal associations between multiple circulating metabolites, GM taxa and immune phenotypes and DIE susceptibility, including risk-increasing bile acid-related and acylcarnitine species, specific bacterial taxa, and monocytic/dendritic-cell traits, and protective lipid species, short-chain-fatty-acid-linked genera and CD45RA Conclusion: This multi-omics MR framework supports a putative gut-immune-pelvic pathway in DIE and identifies a biologically plausible five-gene tissue-level signature consistent with lesion-associated fibrotic and immune-inflammatory remodeling.