This is the repository of codes for the paper entitled "Predicting a Molecular Fingerprint from an Electron Ionization Mass Spectrum with Deep Neural Networks" (DOI: 10.1021/acs.analchem.0c01450).
We exported each mass spectrum file (msp file) and molecular file (sdf) from NIST 2017 manually and save to db file (see scripts/NIST2DB.py for codes). When training, we retrieved information from the db file. However, this repository only contains the source codes without any data or pretrained models, due to the models were trained by NIST dataset.
DeepEI contain two main parts: 1. Predicting molecular fingerprint from EI-MS (Fingerprint folder); 2. Predicting retention index from structure (retention folder). Each folder contains the codes for data pretreatment, model training and model selection.
Moreover, the scripts folder contains the scripts for convert NIST database to numpy object; the Discussion folder contains the scripts for evaluating the identification performance, and comparing with NEIMS package. The corresponding results are also included.
Contact: ji.hongchao@foxmail.com