TPGPred: A Mixed-Feature-Driven Approach for Identifying Thermophilic Proteins Based on GradientBoosting.

Area of Science:

• Biochemistry
• Bioinformatics
• Machine Learning

Background:

• Thermophilic proteins are vital for high-temperature biological processes and industrial applications due to their stability.
• Predicting thermophilic proteins aids in fundamental research and biotechnological development.

Purpose of the Study:

• To develop a machine learning model for accurate prediction of thermophilic proteins.
• To identify optimal feature engineering and classification methods for thermophilic protein prediction.

Main Methods:

• Developed a GradientBoosting prediction model, TPGPred.
• Utilized a large-scale dataset of thermophilic and non-thermophilic protein sequences.
• Employed feature-engineering and systematic evaluation of machine learning algorithms.

Main Results:

• TPGPred achieved an accuracy score > 0.95 and an AUROC score > 0.98 on an independent test set.
• The model was trained on a dataset of 5652 protein sequences.
• Optimal feature combinations and classification models were identified.

Conclusions:

• TPGPred provides a reliable method for identifying and classifying thermophilic proteins.
• The findings offer insights into thermophilic protein characteristics.
• This work lays a foundation for the industrial application of thermophilic proteins.