MEDUSA: Prediction of Protein Flexibility from Sequence

Area of Science:

• Structural biology
• Computational biology
• Bioinformatics

Background:

• Protein flexibility is crucial for understanding molecular mechanisms like stability, interactions, and function.
• B-factor from X-ray crystallography is a common, but limited, measure of protein flexibility.
• A growing gap exists between resolved protein structures and available sequences, necessitating computational prediction methods.

Purpose of the Study:

• To develop a computational tool for predicting protein flexibility directly from amino acid sequences.
• To provide an accessible resource for researchers studying protein dynamics.

Main Methods:

• Developed MEDUSA, a Deep Learning-based tool utilizing convolutional neural networks.
• Input features include evolutionary information from homologous sequences and amino acid physico-chemical properties.
• Trained on a non-redundant dataset of X-ray crystallography structures.

Main Results:

• MEDUSA accurately predicts protein flexibility in two, three, and five classes.
• The tool identifies potentially highly deformable protein regions.
• Provides insights into the general dynamic properties of proteins.

Conclusions:

• MEDUSA offers a valuable computational approach for protein flexibility prediction.
• The tool enhances the study of protein stability, interactions, and function.
• Freely available as a web-server and standalone utility for broad accessibility.