metadata

dataset_info:
  features:
    - name: input_ids
      sequence: int32
    - name: coords
      sequence:
        sequence: float64
    - name: labels
      sequence: float64
  splits:
    - name: train
      num_bytes: 1614130168
      num_examples: 25400
    - name: val
      num_bytes: 192093760
      num_examples: 2800
    - name: test
      num_bytes: 1264620360
      num_examples: 16014
  download_size: 1497041025
  dataset_size: 3070844288
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: val
        path: data/val-*
      - split: test
        path: data/test-*

PSR: Protein Structure Ranking

Overview

This task relates to predicting the three-dimensional structure of a protein molecule, given its sequence. A total of around 700 protein targets are included, which consist of protein targets from Critical Assessment of Structure Prediction (CASP) 5-13.

We phrase this problem as decoy ranking. For each protein target, we include its decoys sets compiled from the CASP decoy sets released for the Model Quality Assessment (MQA). We relax those structures with the SCWRL4 software (Krivov and Dunbrack, 2019) to improve side-chain conformations. For each decoy, we calculate the RMSD, TM-score, GDT_TS, and GDT_HA scores to the experimentally determined structure using the TM-score software (Zhang and Skolnick, 2007).

Datasets

splits:
- split-by-year: This dataset contains data in the CASP dataset, but is split temporally. More specifically, we randomly split the targets in CASP5-10 and randomly sample 50 decoys for each target to generate the training and validation sets (508 targets for training, 56 targets for validation), and use the CASP11 Stage 2 as test set (85 targets total, with 150 decoys for each target, excluding the native structures).

Citation Information

@article{townshend2020atom3d,
  title={Atom3d: Tasks on molecules in three dimensions},
  author={Townshend, Raphael JL and V{\"o}gele, Martin and Suriana, Patricia and Derry, Alexander and Powers, Alexander and Laloudakis, Yianni and Balachandar, Sidhika and Jing, Bowen and Anderson, Brandon and Eismann, Stephan and others},
  journal={arXiv preprint arXiv:2012.04035},
  year={2020}
}

@article{kryshtafovych2019critical,
  title={Critical assessment of methods of protein structure prediction (CASP)—Round XIII},
  author={Kryshtafovych, Andriy and Schwede, Torsten and Topf, Maya and Fidelis, Krzysztof and Moult, John},
  journal={Proteins: Structure, Function, and Bioinformatics},
  volume={87},
  number={12},
  pages={1011--1020},
  year={2019},
  publisher={Wiley Online Library}
}