Identifying emphysema risk using brominated flame retardants exposure: a machine learning predictive model based on...

Area of Science:

• Environmental Health
• Pulmonary Medicine
• Computational Biology

Background:

• Emphysema significantly contributes to lung disease progression and associated health risks.
• Environmental exposures, like brominated flame retardants (BFRs), are potential emphysema risk factors but are understudied.
• The role of BFRs in emphysema prediction remains largely overlooked.

Purpose of the Study:

• To develop a machine learning (ML) model for predicting emphysema risk.
• To incorporate BFRs exposure data and demographic characteristics into the predictive model.
• To assess the influence of BFRs on emphysema risk prediction.

Main Methods:

• Utilized the NHANES (2005-2016) dataset with 8,205 participants.
• Applied eight ML algorithms (lightGBM, MLP, DT, KNN, RF, SVM, Enet, XGBoost) on training (70%) and testing (30%) sets.
• Employed SHAP and Partial Dependence Plots (PDP) for model interpretability.

Main Results:

• The MLP model achieved the highest performance with an AUC of 0.83.
• Age and PBB153 were identified as the most influential predictors.
• Higher BFRs exposure, especially PBB153, showed a strong association with increased emphysema risk, confirmed by SHAP and WQS analyses.

Conclusions:

• BFR exposure holds significant predictive value for emphysema risk assessment.
• Incorporating environmental factors like BFRs is crucial for accurate disease prediction models.
• Findings support integrating BFR assessment into personalized health risk evaluations and public health strategies.