MetGem Software for the Generation of Molecular Networks Based on the t-SNE Algorithm

Area of Science:

• Bioinformatics
• Metabolomics
• Computational Chemistry

Background:

• Molecular networking (MN) is a key bioinformatics tool in metabolomics, relying on MS2 fragmentation similarity for compound analysis.
• Current MN workflows involve filtering similarity scores, which can be time-consuming and lead to information loss.
• Dimensionality reduction algorithms like t-distributed stochastic neighbor embedding (t-SNE) offer alternative approaches for data representation.

Purpose of the Study:

• To develop a novel software, MetGem, that overcomes limitations in existing molecular networking workflows.
• To integrate traditional MN with t-SNE for a more comprehensive analysis of MS2 spectral data.
• To provide a user-friendly, efficient tool for MS2 spectral comparison and network generation.

Main Methods:

• Developed MetGem, a local software integrating GNPS-style molecular networking with t-distributed stochastic neighbor embedding (t-SNE).
• Implemented parallel investigation of both MN and t-SNE representations of raw MS2 spectral data.
• Enabled real-time parameter tuning and rapid network generation for small datasets.

Main Results:

• MetGem offers two complementary data representations: MN for clear cluster separation and t-SNE for preserving spectral interactions.
• The software allows for dynamic parameter adjustment and quick network creation.
• A unified, user-friendly interface was created for efficient MS2 comparison and spectral network generation.

Conclusions:

• MetGem provides a significant advancement in metabolomic data analysis by combining MN and t-SNE.
• The software addresses the need for a dedicated, efficient, and user-friendly tool for spectral network analysis.
• MetGem facilitates a more thorough understanding of MS2 spectral datasets through its integrated approach.