ESMStabP: A Regression Model for Protein Thermostability Prediction.

Area of Science:

• Computational Biology
• Biotechnology
• Protein Science

Background:

• Accurate prediction of protein thermostability is vital for biotechnology, pharmaceuticals, and food science.
• Experimental determination of protein melting temperatures is resource-intensive.
• There is a need for efficient computational methods to predict protein stability.

Purpose of the Study:

• To introduce ESMStabP, an enhanced regression model for predicting protein thermostability.
• To improve upon existing computational models for protein thermostability prediction.
• To leverage advanced protein language models and curated datasets for enhanced predictive accuracy.

Main Methods:

• Assembled a larger, cleaned dataset by combining existing thermostability datasets.
• Developed ESMStabP, building upon the DeepStabP model.
• Incorporated embeddings from the ESM2 protein language model and thermophilic classifications.

Main Results:

• ESMStabP significantly outperforms DeepStabP and other existing predictors.
• Achieved a high R-squared (R²) of 0.95.
• Attained a Pearson correlation coefficient (PCC) of 0.97, indicating strong predictive performance.

Conclusions:

• ESMStabP represents a significant advancement in computational protein thermostability prediction.
• Highlights the importance of dataset quality and specific model layer identification.
• Suggests future research directions for improving protein stability prediction models.