pae_to_domains

Graph-based community clustering approach to extract protein domains from a predicted aligned error matrix

Overview

Using a predicted aligned error matrix corresponding to an AlphaFold2 model (e.g. as downloaded from https://alphafold.ebi.ac.uk/), returns a series of lists of residue indices, where each list corresponds to a set of residues clustering together into a pseudo-rigid domain.

Requirements

• Python >=3.7
• NetworkX >= 2.6.2

Known Issues

Due to an internal implementation issue in NetworkX (Issue #4992) some combinations of PAE matrix and resolution can lead to a KeyError. Solutions to this are being explored, and it will hopefully be fixed in the next NetworkX release.

Usage

While primarily intended as a code snippet to be incorporated into larger projects, this can also be called from the command line. At its simplest:

python pae_to_domains.py pae_file.json

... will yield a .csv file with each line providing the indices for one residue cluster. Full help for the command-line version:

positional arguments:
  pae_file              Name of the PAE JSON file.

optional arguments:
  -h, --help            show this help message and exit
  --output_file OUTPUT_FILE
                        Name of output file (comma-delimited text format.
                        Default: clusters.csv
  --pae_power PAE_POWER
                        Graph edges will be weighted as 1/pae**pae_power.
                        Default: 1.0
  --pae_cutoff PAE_CUTOFF
                        Graph edges will only be created for residue pairs
                        with pae