AlphaKnot: server to analyze entanglement in structures predicted by AlphaFold methods

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Protein structure prediction accuracy has significantly improved with tools like AlphaFold.
• Understanding protein entanglement is crucial for deciphering protein function and evolution.
• Quantifying and classifying protein knotting in predicted structures remains a challenge.

Purpose of the Study:

• To introduce AlphaKnot, a computational server for analyzing and visualizing protein entanglement in AlphaFold models.
• To provide a probabilistic framework for defining and categorizing protein knotting.
• To create a database of protein knotting information derived from publicly available AlphaFold predictions.

Main Methods:

• Development of a web server integrating AlphaFold model analysis with pLDDT confidence values.
• Probabilistic definition of protein knotting and knot core identification.
• Classification of entanglements into 'Knots', 'Unsure', and 'Artifacts' based on confidence scores.

Main Results:

• AlphaKnot provides a user-friendly web interface for analyzing knotting in individual AlphaFold predictions.
• A comprehensive database of protein knotting for 21 proteomes is now available.
• Knot types, complexities, and core locations are visualized using matrix diagrams.

Conclusions:

• AlphaKnot offers valuable insights into protein entanglement, aiding in the interpretation of AlphaFold models.
• The server and database facilitate studies on entanglement-function correlations, protein evolution, and modeling.
• Probabilistic assessment of knotting, informed by pLDDT, enhances the reliability of entanglement analysis.