owenustc / deepaccnet-msa

ErrorPredictor-MSA.py

A script for predicting protein model accuracy.

usage: ErrorPredictorMSA.py [-h] [--pdb] [--leavetemp] [--verbose] [--process PROCESS] [--gpu GPU] [--featurize]
                            [--reprocess] [--ensemble]
                            distogram infolder ...

Error predictor network with predicted distogram

positional arguments:
  distogram             predicted distogram from trRosetta (npz format, key for distogram should be 'dist')
  infolder              input folder name full of decoy pdbs having same sequence or path to a single pdb
  outfolder             output folder name. If a pdb path is passed this needs to be a .npz file. Can also be empty.
                        Default is current folder or pdbname.npz

optional arguments:
  -h, --help            show this help message and exit
  --pdb, -pdb           Running on a single pdb file instead of a folder (Default: False)
  --leavetemp, -lt      leaving temporary files (Default: False)
  --verbose, -v         verbose flag (Default: False)
  --process PROCESS, -p PROCESS
                        # of cpus to use for featurization (Default: 1)
  --gpu GPU, -g GPU     gpu device to use (default gpu0)
  --featurize, -f       running only featurization (Default: False)
  --reprocess, -r       reprocessing all feature files (Default: False)
  --ensemble, -e        ensemble (Default: False)

v0.0.1

Required softwares

• Python3.5>
• Pyrosetta, trRosetta
• Tensorflow 1.14 (not Tensorflow 2.0)

How to look at outputs

Output of the network is written to [input_file_name].npz. You can extract the predictions as follows.

import numpy as np

x = np.load("testoutput.npz")

lddt = x["lddt"]           # per residue lddt
estogram = x["estogram"]   # per pairwise distance e-stogram
mask = x["mask"]           # mask predicting native < 15

Perhaps lddt is the easiest place to start as it is per-residue quality score. You can simply take an average if you want a global score per protein structure.

Trouble shooting

• If ErrorPredictor.py returns an OOM (out of memory) error, your protein is probably too big. Try getting on titan instead of rtx2080 or run without gpu if running time is not your problem. You can also truncate your protein structures although it is not recommended.
• If you get an import error for pyErrorPred, you probably moved the script out of LocalAccuacyPredictor. In that case, you would have to add pyErrorPred to python path or do so within the script.
• Send an e-mail at hiranumn at cs dot washington dot edu.

Updates

• Repo initialized 2020.11.3. (Planning to merge with DeepAccNet soon).