Using Environmental Mixture Exposure-Triggered Biological Knowledge-Driven Machine Learning to Predict Early...

Area of Science:

• Environmental epidemiology
• Reproductive toxicology
• Bioinformatics

Background:

• Assessing environmental mixture effects on reproductive health is challenging.
• Knowledge graph networks (KGNs) show promise for biological interpretation but are underutilized in epidemiology, especially with small sample sizes.

Purpose of the Study:

• To develop and validate a novel framework integrating biological knowledge graph-based networks (BKGNs) and machine learning (ML) for predicting early pregnancy loss (EPL).
• To identify critical environmental exposures and biological pathways associated with EPL in women undergoing in vitro fertilization and embryo transfer (IVF-ET).

Main Methods:

• Recruited 116 women undergoing IVF-ET, collecting clinical data and biological samples (hair, serum, follicular).
• Measured 16 per- and polyfluoroalkyl substances (PFAS) and 41 metal(loid)s.
• Developed a framework coupling BKGNs (integrating Gene Ontology and protein pathways) with ML to predict EPL, testing its robustness with reduced sample sizes.

Main Results:

• The Gene Ontology-integrated BKGN-ML model achieved a high area under the curve (AUC) of 0.876, outperforming other models (AUC = 0.819).
• The model remained effective even with a 40% reduction in sample size.
• Identified key exposures like serum selenium and chromium, and biological perturbations such as cell population proliferation and apoptotic nuclear changes linked to EPL.

Conclusions:

• The proposed BKGN-ML framework is a robust, cost-effective tool for predicting exposure-associated reproductive health outcomes.
• This mechanistic approach provides valuable insights into the links between environmental mixture exposures and early pregnancy loss.