VirMolAnalyte: An AI-Driven In Silico Metabolite Annotation Tool

Area of Science:

• Metabolomics
• Drug Discovery
• Spectroscopy

Background:

• Metabolites are vital for biological functions and drug development.
• Nuclear Magnetic Resonance (NMR) spectroscopy is key for metabolite structure identification.
• Current NMR spectral annotation requires expensive experimental data.

Purpose of the Study:

• To introduce VirMolAnalyte, a tool for analyzing 13C DEPT NMR spectra without experimental data.
• To enable high-quality structural analysis of metabolites computationally.
• To overcome the limitations of experimental data dependency in NMR analysis.

Main Methods:

• VirMolAnalyte uses a deep neural network to extract molecular fingerprints from 13C DEPT NMR spectra.
• A filter-evaluator strategy is employed for multidimensional in silico database searching.
• The tool processes NMR spectral data computationally.

Main Results:

• The filter-evaluator strategy achieved 94.2% Top 1 accuracy on DB6123, surpassing traditional methods.
• VirMolAnalyte showed 90.0% Top 1 and 100.0% Top 5 accuracy on a large in silico metabolite database.
• The tool demonstrated high accuracy in metabolite identification from NMR spectra.

Conclusions:

• VirMolAnalyte offers an innovative computational approach for metabolite analysis using NMR data.
• The tool significantly advances new drug discovery and metabolomics research.
• VirMolAnalyte reduces reliance on costly experimental data for NMR spectral annotation.