Predicting Protein Interactions with BteA in Bordetella pertussis Pathogenesis Using Machine Learning

This study employed machine learning techniques to identify protein interactions with BteA in Bordetella pertussis, aiming to enhance our understanding of whooping cough pathogenesis and discover potential targets for new drug.Whooping cough (pertussis), caused by B. pertussis, remains a significant global health challenge despite widespread vaccine availability.B. pertussis protein BteA, delivered into host cells through the type III secretion system, plays a crucial role in promoting host cell death and pathogenesis.We utilized machine learning to predict proteins that interact with BteA.First, the STRING database was used to learn protein interactions, and Biopython was employed to extract four biological features for each protein: isoelectric point, total number of positively and negatively charged amino acid residues, and GRAVY.Six machine learning models, including XGBoost, Random Forest, and LightGBM, were trained using these features.The top models were applied to predict proteins likely to interact with BteA.RpoB, RpID, BP2755 and RpsK were identified with high interaction probability.Feature importance analysis using Shapley values showed that isoelectric points and GRAVY contributed significantly to the prediction models.Our findings demonstrate the effectiveness of machine learning in predicting protein-protein interactions within B. pertussis, providing valuable insights into the interaction network of BteA.The identified proteins offer promising targets for new drug, potentially