Obesity and Constipation: An Integrative Computational Genomics Approach Using Two-Sample Mendelian Randomization

The causal link between obesity and constipation remains unclear, with conventional genetic analyses relying merely on statistical indices lacking multi-omics evidence, a key challenge for biomedical multimodal foundation-model research. Using summary-level GWAS data from GIANT, UK Biobank and FinnGen, we performed bidirectional two-sample Mendelian randomization, multivariable MR and colocalization, complemented by an XGBoost-based multimodal AI framework for variant prioritization. Genetically predicted higher BMI (OR=1.200, P=1.921×10 -4 ) and WHR (OR=1.270, P=2.672×10 -4 ) causally increased constipation risk, whereas reverse causation was absent. Two trait-specific shared loci were identified: rs543874 for BMI-constipation and rs1936789 for WHRadjBMI-constipation. Abdominal fat distribution exerted stronger pathogenic effects. This work integrates multimodal AI with genetic causal inference, offering insights for stratified constipation prevention and target discovery.