TEM-1 is a β-lactamase that hydrolyzes penicillin as an antibiotic. The catalytic mechanism of the TEM-1 penicillin-hydrolysis has been well studied to understand antibiotic resistance related to β-lactamases. However, the recognition of ligands in different states by TEM-1 and β-lactamases in general is less understood but important for antibiotic resistance study. It has been shown that the binding between TEM-1 with penicillin during different stages of hydrolysis plays important role for TEM-1’s penicillin-hydrolysis functions. To shed light on the contribution of ligand binding and recognition by TEM-1 for its function, we developed machine learning classification model using random forest method for three states of TEM-1 binding benzyl-penicillin system: reactant state (TEM-1 β-lactamase binding with benzyl penicillin), product state (TEM-1 β-lactamase with a hydrolyzed benzyl penicillin) and apo state (TEM-1 β-lactamase without ligand). Interestingly, this classification model could differentiate the reactant and apo states well from the product state, but not between the reactant and apo states, suggesting that the hydrolyzed product of penicillin is a unique ligand binding state of TEM-1. Based on the importance related to individual residues available from the classification model, key structural features of TEM-1 with regard to ligand binding and recognition are determined, including secondary structures surrounding the active site of TEM-1. In summary, machine learning methods are demonstrated as effective tools to analyze the dynamics of proteins and provide insight into the contribution of the key residues and secondary structures of proteins to their functions.

Machine learning-based classification model for β-Lactamase recognizing its ligand: TEM-1 and benzyl-penicillin complex

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.