NMR-TS

NMR-TS integrates de novo molecule generation, deep learning, and density functional theory (DFT) to identify candidate molecular structures that match target NMR spectra for chemical structure elucidation.

Key Features:

• Automated Identification: Uses deep learning models to compare computed NMR spectra to a target spectrum and identify candidate molecules.
• De Novo Molecule Generation: Generates candidate molecular structures de novo to propose molecules not present in existing databases.
• Integration of Deep Learning and DFT: Combines deep learning with DFT-computed NMR spectra and performs extensive computational runs, averaging 5451 DFT calculations per spectrum.
• Proof-of-Concept Results: In a proof-of-concept application, identified six of nine prototypical metabolites from computed spectra and produced proximal molecule matches when exact matches were not found.

Scientific Applications:

• Metabolomics: Identification of metabolites from NMR spectra for metabolomics analyses.
• Drug Discovery: Exploration of chemical space and proposal of candidate structures for unknown compounds in drug discovery efforts.
• Chemical Structure Elucidation: Elucidation of molecular structures from NMR spectra and exploration of proximal molecular candidates to support studies of biochemical pathways.

Methodology:

De novo generation of candidate molecules, DFT computation of each candidate's NMR spectrum, and comparison of computed spectra to the target using deep learning models.