mdCATH: A Large-Scale MD Dataset for Data-Driven Computational Biophysics

This dataset comprises all-atom systems for 5,398 CATH domains, modeled with a state-of-the-art classical force field, and simulated in five replicates each at five temperatures from 320 K to 450 K.

Availability

• torchmd-net dataloader
• playmolecule
• scripts to load, convert and rebuild

Citing The Dataset

Please cite this manuscript for papers that use the mdCATH dataset:

Mirarchi, A., Giorgino, T. & De Fabritiis, G. mdCATH: A Large-Scale MD Dataset for Data-Driven Computational Biophysics. Sci Data 11, 1299 (2024). https://doi.org/10.1038/s41597-024-04140-z. Preprint available at arXiv:2407.14794 (2024).

Dataset Size

Description	Value
Domains	5,398
Trajectories	134,950
Total sampled time	62.6 ms
Total atoms	11,671,592
Total amino acids	740,813
Avg. traj. length	464 ns
Avg. system size	2,162 atoms
Avg. domain length	137 AAs
Total file size	3.3 TB