Predicting nutrition and environmental factors associated with female reproductive disorders using a knowledge graph and random forests

• DOI: 10.1016/j.ijmedinf.2024.105461
• Publication Date: 2024-04-17T15:59:12Z
• Authors: Lauren E Chan; Elena Casiraghi; Justin Reese; Quaker E. Harmon; Kevin Schaper; Harshad Hegde; Giorgio Valentini; Charles Schmitt; Alison Motsinger-...; Janet E Hall; Christopher J Mun...; Peter N Robinson; Melissa A Haendel
• Keywords: Adult; Logistic Models; Random Forest; JGM; Genital Diseases; 610; Humans; Nutritional Status; Female; Environmental Exposure; Genital Diseases, Female; Article; Diet

Female reproductive disorders (FRDs) are common health conditions that may present with significant symptoms. Diet and environment potential areas for FRD interventions. We utilized a knowledge graph (KG) method to predict factors associated FRDs (for example, endometriosis, ovarian cyst, uterine fibroids). harmonized survey data from the Personalized Environment Genes Study (PEGS) on internal external environmental exposures biomedical ontology content. merged ontologies supplemental nutrient agricultural chemical create KG. analyzed KG by embedding edges applying random forest edge prediction identify variables potentially FRDs. also conducted logistic regression analysis comparison. Across 9765 PEGS respondents, resulted in 8535 or suggestive predicted links between chemicals, phenotypes, diseases. Amongst these links, 32 were exact matches when compared results, including comorbidities, medications, foods, occupational exposures. Mechanistic underpinnings of documented literature support some our findings. Our methods useful predicting possible associations large, survey-based datasets added information directionality magnitude effect regression. These results should not be construed as causal but can hypothesis generation. This investigation enabled generation hypotheses variety Future investigations prospectively evaluate hypothesized impact