Public Data files for MassFormer

Public data files for experiments in MassFormer. See the Github repository for instructions on how to use this data.

Raw Data:

casmi_2016.tgz - Critical Assessment of Small Molecules 2016, used for model evaluation.

mb_na_msms.msp.gz - MassBank of North America export of LC-MS/MS spectra, used for model evaluation.

cid_smiles.tsv.gz - Mapping of CID to SMILES strings, obtained from PubChem.

Processed Data:

casmi.tgz - Processed spectrum and molecule data for the CASMI benchmark.

pcasmi.tgz - Processed spectrum and molecule data for the pseudo-CASMI (pCASMI) benchmark.

proc_demo.tgz - Processed spectrum and molecule data for the demo (refer to code repository for more information).

Model Checkpoints:

demo.pkl - Checkpoint of a MassFormer model trained on MoNA data, for the purposes of running the demo.

checkpoint_best_pcqm4mv2.pt - Checkpoint of a Graphormer model pretrained on the PCQM4M dataset, used for initialization of some MassFormer models. Copied from this url. Please refer to the Graphormer repository for more information.