Artificial Intelligence-Powered Construction of a Microbial Optimal Growth Temperature Database and Its Impact on Enzyme Optimal Temperature Prediction

Abstract

Accurate prediction of enzyme optimal temperature (Topt) is crucial for identifying enzymes suitable for catalytic functions under extreme bioprocessing conditions. The optimal growth temperature (OGT) of microorganisms serves as a key indicator for estimating enzyme Topt, reflecting an evolutionary temperature balance between enzyme-catalyzed reactions and the organism's growth environments. Existing OGT databases, collected from culture collection centers, often fall short as culture temperature does not precisely represent the OGT. Models trained on such databases yield inadequate accuracy in enzyme Topt prediction, underscoring the need for a high-quality OGT database. Herein, we developed AI-based models to extract the OGT information from the scientific literature, constructing a comprehensive OGT database with 1155 unique organisms and 2142 OGT values. The top-performing model, BioLinkBERT, demonstrated exceptional information extraction ability with an EM score of 91.00 and an F1 score of 91.91 for OGT. Notably, applying this OGT database in enzyme Topt prediction achieved an R 2 value of 0.698, outperforming the R 2 value of 0.686 obtained using culture temperature. This emphasizes the superiority of the OGT database in predicting the enzyme Topt and underscores its pivotal role in identifying enzymes with optimal catalytic temperatures.