An End-to-end Deep Learning Framework For Translating Mass Spectra To De-novo Molecules

This article presents Spec2Mol, a deep learning architecture for molecular structure recommendation given mass spectra alone. Spec2Mol is inspired by Speech2Text deep learning architectures and is based on an encoder-decoder architecture. The encoder learns the spectra embeddings, while the decoder, pre-trained on a massive dataset of chemical structures, reconstructs SMILES sequences of the recommended chemical structures. The evaluation of Spec2Mol showed that it is able to identify the presence of key molecular substructures from its mass spectrum, and shows on par performance when compared to existing fragmentation tree methods.