Protein topology from predicted residue contacts

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• Correlated positions in multiple sequence alignments (MSAs) predict residue contacts.
• These predictions are often sparse and uncertain, limiting their application.
• Current methods struggle with larger protein structures.

Purpose of the Study:

• To develop a robust method for predicting residue contacts in proteins.
• To overcome limitations of sparse and uncertain contact predictions from MSAs.
• To enable accurate structure prediction for larger proteins using sequence data alone.

Main Methods:

• Grouping predicted residue contacts by secondary structure elements.
• Enumerating possible packing arrangements of these elements combinatorially.
• Down-weighting inconsistent interactions and up-weighting missing interactions.

Main Results:

• Achieved improved consistency in predicted protein-protein interactions.
• Successfully applied the method to predict structures of proteins up to 200 residues.
• Demonstrated a method applicable to larger protein structures than previously possible using sequence data.

Conclusions:

• Grouping contacts by secondary structure elements significantly enhances prediction accuracy.
• The developed method overcomes limitations of sparse MSA data for protein structure prediction.
• This approach enables reliable prediction of larger protein structures from sequence information.