Sequence-Based Prediction for Protein Solvent Accessibility

Area of Science:

• Structural biology
• Computational biology
• Biochemistry

Background:

• Amino acid positioning (surface vs. core) in folded proteins dictates environmental interaction.
• Residue accessibility is vital for protein functions like enzymatic activity, allostery, multimerization, binding, and immunogenicity.

Purpose of the Study:

• To develop a novel sequence-based predictor for amino acid accessibility.
• To provide insights into protein function and residue relevance through solubility prediction.

Main Methods:

• Feature extraction from three-dimensional protein structures.
• Testing various machine learning algorithms.
• Utilizing the long short-term memory (LSTM) deep learning method for prediction.

Main Results:

• The LSTM deep learning method showed the best performance among tested algorithms.
• The developed SolAcc tool demonstrated superior performance in blind tests compared to state-of-the-art predictors.

Conclusions:

• SolAcc is a novel, freely available tool for predicting amino acid accessibility.
• The tool offers accurate predictions based on sequence data, aiding in understanding protein structure-function relationships.