Machine Learning Approaches for Differentiating Thermophilic and Mesophilic Lipases

Differentiating thermophilic proteins from their mesophilic counterparts presents a significant challenge, yet achieving this distinction is crucial for the rational design of more stable proteins. In this study, a systematic analysis was performed on 3,715 unreviewed bacterial lipase enzymes obtained from the UniProt web server and screened according to their Tm values. Furthermore, a tree was constructed using the MEGA 11 program and lipase sequences from different families were selected. The final dataset consists of 88 mesophilic proteins and 123 thermophilic proteins were used. We found that Ile, Leu, aliphatic index, hydropathy, aliphatic amino acids, hydrophobic amino acids, tiny amino acids, and small amino acids are the key variables distinguishing thermophilic from mesophilic lipase proteins. These findings suggest that amino acid composition is crucial in differentiating these two groups.