Predicting the Real-Valued Inter-Residue Distances for Proteins.

Area of Science:

• Biophysics
• Computational Biology
• Structural Biology

Background:

• Protein structure prediction from amino acid sequences is a fundamental challenge in biophysics.
• Current methods using inter-residue contact prediction are approaching their performance limits.
• Existing distance prediction methods simplify real-valued distances into classification problems.

Purpose of the Study:

• To develop a lightweight, regression-based method for predicting continuous, real-valued inter-residue distances.
• To improve the accuracy and speed of protein structure prediction.
• To enable direct structure prediction for membrane proteins.

Main Methods:

• Utilized a generative adversarial network (GAN) to model geometric relationships between residue pairs.
• Developed a regression-based approach for predicting continuous inter-residue distances.
• Integrated the predicted distance map with the CNS suite for rapid structure modeling.

Main Results:

• The method accurately predicts continuous, real-valued inter-residue distances.
• Generated protein models achieve quality comparable to state-of-the-art methods on CASP13 targets.
• Demonstrated direct applicability to membrane protein structure prediction without transfer learning.

Conclusions:

• The proposed generative adversarial network offers a rapid and effective approach for protein structure prediction.
• This regression-based method overcomes limitations of previous classification-based distance predictions.
• The technique shows promise for diverse protein types, including membrane proteins.