Optimal subset selection of primary sequence features using the genetic algorithm for thermophilic proteins...

Area of Science:

• Computational Biology
• Protein Science
• Bioinformatics

Background:

• Thermophilic proteins exhibit enhanced stability at high temperatures.
• Identifying features that confer thermostability is crucial for protein engineering.
• Existing methods may require extensive feature selection for accurate classification.

Purpose of the Study:

• To develop and evaluate a computational method for distinguishing thermophilic from non-thermophilic proteins.
• To identify key amino acid and dipeptide features associated with protein thermostability.
• To provide a tool for aiding in the design of more stable proteins.

Main Methods:

• A hybrid approach combining a genetic algorithm (GA) with multiple linear regression (MLR) was employed.
• The method utilized features derived from amino acids and g-gap dipeptides.
• Training and validation were performed on a benchmark dataset of 915 thermophilic and 793 non-thermophilic proteins.

Main Results:

• The GA-MLR method achieved a high overall accuracy of 95.4% in a Jackknife test.
• Accuracy varied between 85.8% and 96.9% depending on protein size.
• Independent tests confirmed the method's robustness with accuracies of 93%, 93.4%, and 91.8%.

Conclusions:

• The GA-MLR approach is a powerful tool for feature selection in identifying thermostable proteins.
• The identified features can inform the design of proteins with enhanced stability.
• This computational strategy offers a significant aid in protein engineering for thermostability.