MQAB4AF2

Model Quality Assessment (Estimation of Model Accuracy) benchmarking for AlphaFold2 structures

Dataset generation

Requirement

python=3.9.2
requests=2.26.0
numpy=1.21.3
pandas=1.3.4
tqdm=4.62.3
prody=2.0
joblib=1.1.0
dssp=3.0.0
pytest=6.2.5

Procedure

1. Target selection
2. Run AlphaFold2
3. [Optional] Run MQA methods

Download dataset

You can download AlphaFold2 structure data for 500 protein sequences from http://www.cb.cs.titech.ac.jp/af2/af2_500_targets.tar.gz (5.8GB).

The dataset includes protein target sequences, native and predicted structures of the targets, and labels. For more information, see here.

Reference

Yuma Takei and Takashi Ishida, "How to select the best model from AlphaFold2 structures?", bioRxiv, 2022. Available from https://doi.org/10.1101/2022.04.05.487218. (Supporting information is available here.)

Acknowledgement

AlphaFold v2.0.1, ColabFold, and LocalColabFold v1.0.0 were used to predict structures.

• Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, Tunyasuvunakool K, Bates R, Žídek A, Potapenko A and Bridgland A. "Highly accurate protein structure prediction with AlphaFold."
  Nature. 2021 Aug;596(7873):583-9. doi: 10.1038/s41586-021-03819-2

• Mirdita M, Schütze K, Moriwaki Y, Heo L, Ovchinnikov S and Steinegger M. "ColabFold - Making protein folding accessible to all."
  bioRxiv (2021) doi: 10.1101/2021.08.15.456425

• Moriwaki Y. "LocalColabFold" Available from: https://github.com/YoshitakaMo/localcolabfold.