Enzyme Turnover Number Prediction Based on Protein 3D Structures

Protein function prediction has long been a widely discussed task in the field of synthetic biology, and it is of paramount importance for gaining a deeper understanding of the roles and interactions of proteins within living organisms. Since the 3D structure data of proteins obtained experimentally are far less in quantity than the corresponding protein sequence data, most experiments related to protein function prediction currently rely on using protein sequences as training data, although 3D protein structures contain much more information. Here, an enzyme turnover number prediction model (PSKcat) is proposed based on 3D protein structures. PSKcat takes protein PDB files as input, represents proteins using a modified pre-trained model called GearNet-Edge for 3D protein structures, and combines graph neural network to characterize the substrates involved in enzyme reactions. In order to verify the effectiveness of the model, several enzyme reaction datasets were constructed, and multiple groups of comparative experiments were conducted. The experimental results demonstrate the feasibility of using 3D protein structures for enzyme function prediction, which opens up avenues for further exploration of the applications of 3D protein structures in the future.