metadata

dataset_info:
  features:
    - name: input_ids
      sequence: int32
    - name: coords
      sequence:
        sequence: float64
    - name: labels
      dtype: int64
    - name: token_type_ids
      sequence: int8
  splits:
    - name: train
      num_bytes: 1419748993
      num_examples: 2864
    - name: val
      num_bytes: 394945349
      num_examples: 937
    - name: test
      num_bytes: 214263730
      num_examples: 347
  download_size: 774980058
  dataset_size: 2028958072
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: val
        path: data/val-*
      - split: test
        path: data/test-*

Mutation Stability Prediction

Overview

The Mutation Stability Prediction (MSP) task involves classifying whether mutations in the SKEMPI 2.0 database (J. Jankauskaite, B. Jiménez-García et al., 2019) are stabilizing or not using the provided protein structures.

Each mutation in the MSP task includes a PDB file with the residue of interest transformed to the specified mutant amino acid as well as the native PDB file.

A total of 4148 mutant structures accompanied by their 316 WT structures are provided.

Non-point mutations or mutants that caused non-binding of the complex were screened out from SKEMPI. Additionally, mutations involving a disulfide bond and mutants from the PDBs 1KBH or 1JCK were ignored due to processing difficulties. A label of 1 was assigned to a mutant if the Kd of the mutant protein was less than that of the wild-type protein, indicating better binding, and 0 otherwise. If a mutant had multiple conflicting stability results in the original data, the first result seen when processing the SKEMPI dataset was used.

Datasets

splits:
- split-by-year:

Format

Each item in the dataset contains the following keys:

['input_ids'] The set of atomic numbers for both the mutated and original atoms concatenated together.
['coords'] The 3D coordinates for both the mutated and original atoms concatenated together.
['label'] 1 for better binding, 0 for worse or equal.
['token_type_ids'] A mask that corresponds to which inputs_ids/coords belong to the mutated or original atoms (0 for mutated , 1 for original)

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{jankauskaite2019skempi,
  title={SKEMPI 2.0: an updated benchmark of changes in protein--protein binding energy, kinetics and thermodynamics upon mutation},
  author={Jankauskait{\.e}, Justina and Jim{\'e}nez-Garc{\'\i}a, Brian and Dapk{\=u}nas, Justas and Fern{\'a}ndez-Recio, Juan and Moal, Iain H},
  journal={Bioinformatics},
  volume={35},
  number={3},
  pages={462--469},
  year={2019},
  publisher={Oxford University Press}
}