Assessing a Correlation between Predicted Protein Stability and Antibiotic Resistance of Beta-Lactamase Enzymes

Abstract

Rates of antibiotic resistance in pathogenic bacteria have risen significantly in the past decades, rendering many widely used beta-lactam antibiotics ineffective. The main resistance mechanisms of these bacteria are linked to beta-lactamase enzymes that hydrolyze rings within beta-lactam antibiotics. As a result, predicting the impact of mutations in beta-lactamases can be critical to developing effective antibiotic treatments. The aim of our research is to investigate how antibiotic resistance correlates with protein stability changes. Our research employs computational machine learning models that predict changes in protein stability, including AUTOMUTE 2.0, MAESTROweb, PROST, and MUpro. We compiled a dataset of 83 single residue mutations with known antibiotic resistance degrees. Using these models, we predicted stability scores (ΔΔG) for each mutant in the dataset, as well as classified each mutant by residue depth and secondary structure using Swiss PDB Viewer and PDB database, respectively. Lastly, the analysis of the results indicated that the correlation between the enzyme stability and drug resistance is not statistically significant. Additional features for drug resistance modeling should be considered.