AFragmenter: schema-free, tuneable protein domain segmentation for AlphaFold protein structures

Area of Science:

• Structural Biology
• Bioinformatics
• Computational Biology

Background:

• Protein domain segmentation is essential for understanding protein function, interactions, modeling, and evolution.
• Existing methods often rely on predefined schemes, introducing inconsistencies and biases.
• A new approach is needed for unbiased and flexible protein domain segmentation.

Purpose of the Study:

• To introduce AFragmenter, a novel schema-free tool for protein domain segmentation.
• To leverage network analysis of AlphaFold-predicted structures for improved domain identification.
• To provide users with tuneable parameters for customized segmentation.

Main Methods:

• AFragmenter utilizes network analysis of AlphaFold-predicted protein structures.
• Predicted Aligned Error (PAE) values are used to construct residue networks.
• Leiden clustering algorithm is employed to identify distinct structural domains.
• Users can adjust parameters like contrast threshold and resolution for segmentation control.

Main Results:

• AFragmenter successfully segments protein domains in a schema-free manner.
• The method provides tuneable control over the segmentation process.
• It offers an alternative to traditional, potentially biased, classification schemes.

Conclusions:

• AFragmenter presents a flexible and tuneable approach to protein domain segmentation.
• This method enhances protein modeling and evolutionary studies by providing accurate domain identification.
• The tool is freely available as a Python library and command-line tool.