Roman Bushuiev

Despite great scientific advances, only a tiny fraction of small molecules from nature has been discovered to date. Mass spectrometry is a powerful analytical technique that enables identification of molecules present in biological and environmental samples. However, due to the complexity of its experimental data, existing methods for interpretation of mass spectra are limited and mostly rely on annotated spectral libraries. The goal of the project is to develop novel deep learning models for the identification of molecules directly from experimental mass spectra. More precisely, the project involves research of representation learning from millions of raw, unannotated spectra and subsequent deep generative modeling for de novo elucidation of molecular structures.