Graph-based understanding of isomeric diversity in complex dissolved organic matter

Area of Science:

• Environmental chemistry
• Analytical chemistry
• Biogeochemistry

Background:

• Characterizing isomeric diversity of molecular formulas (MFs) in dissolved organic matter (DOM) is crucial for environmental and biomedical research.
• Ultra-high-resolution mass spectrometry (UHR MS), especially LC-UHR MS, resolves isomeric diversity but faces limitations like cost and complexity.
• Direct infusion UHR MS (DI-UHR MS) is more accessible but lacks chromatographic separation for isomeric resolution.

Purpose of the Study:

• To develop a predictive framework, GUIDE, for inferring isomeric diversity directly from DI-UHR MS data.
• To bridge the gap between the detailed separation of LC-UHR MS and the accessibility of DI-UHR MS.
• To enable high-accuracy prediction of chromatographic isomeric diversity without chromatographic separation.

Main Methods:

• Developed GUIDE, a framework using self-supervised graph learning to represent MFs based on intrinsic features and topological relationships.
• Employed deep neural networks for predicting isomeric diversity from learned MF representations.
• Validated the framework using data from multiple DI-UHR MS platforms (DI-FT-ICR MS, DI-Orbitrap MS).

Main Results:

• GUIDE accurately predicts the chromatographic isomeric diversity of natural DOM.
• The framework demonstrates consistent performance across different DI-UHR MS platforms.
• Achieved resolution of MFs at the isomeric level using DI-UHR MS data.

Conclusions:

• GUIDE provides a powerful, accessible method for resolving isomeric diversity in DOM.
• This approach offers a refined molecular perspective of DOM composition.
• Enables advancements in biogeochemistry, environmental science, and analytical chemistry through detailed molecular insights.