Development of A Machine Learning-Based Predictive Model for Klebsiella pneumoniae Biofilm Formation

Klebsiella pneumoniae is a critical Gram-negative pathogen frequently associated with severe hospital-acquired infections, including pneumonia, urinary tract infections, and bloodstream infections. Its increasing multi-drug resistance, particularly to carbapenems, poses a significant global health challenge. Biofilm formation is a key virulence factor, allowing K. pneumoniae to persist on medical devices and evade antibiotic treatment. This study aimed to develop a predictive model to quantify the individual and synergistic effects of key environmental factors, namely temperature, pH, glucose and sodium chloride (NaCl) concentrations, on K. pneumoniae biofilm formation. To establish the intricate relationships between these environmental parameters and biofilm development, and to construct a robust predictive model, experimental data were analysed using statistical regression and machine learning approaches to construct a predictive model. Among the tested models, XGBoost demonstrated the best predictive performance (R² = 0.6209, Adjusted R² = 0.6209, RMSE = 1.2355, MAE = 0.9384). Feature importance analysis revealed temperature and glucose concentration as the most influential factors, significantly impacting biofilm formation. Partial dependence analysis showed optimal biofilm production was observed at neutral pH and glucose concentrations over 1%, with high temperatures. In addition, high NaCl concentrations significantly stimulated biofilm formation. These findings provide valuable insights into the regulation of K. pneumoniae biofilm formation, which could have practical uses in diverse industrial, medical, and food safety settings.