Coupling Mixed Mode Chromatography/ESI Negative MS Detection with Message-Passing Neural Network Modeling for...

Area of Science:

• Analytical Chemistry
• Metabolomics
• Computational Chemistry

Background:

• Metabolomics research faces challenges in accurately detecting and quantifying polar metabolites.
• Existing methods struggle with the specific retention of simple sugars, sugar phosphates, carboxylic acids, and related amino acids.

Purpose of the Study:

• To develop and validate a novel mixed-mode (MM) chromatography method for improved retention and separation of central carbon metabolism metabolites.
• To build and apply a machine learning model to predict metabolite retention times (RTs) and enhance metabolome coverage.

Main Methods:

• Utilized a mixed-mode chromatography system with a quaternary amine polyvinyl alcohol stationary phase, varying pH, salt concentration, and organic content.
• Coupled the chromatography to a QExactive Orbitrap Mass Spectrometer with negative electrospray ionization (ESI).
• Developed a message-passing neural network (MPNN) to predict RTs from molecular graph features for 398 retained metabolites.

Main Results:

• The MM method successfully separated glucose from fructose and four hexose monophosphates in a single run.
• Assessed 33 metabolite standards, showing good linearity (R² > 0.98 for 30/33), low limits of detection (LLOD < 1 pmole for 26/33), and median CV of 12%.
• The MPNN model achieved predictive RT errors of <2 min for 91% of sugars, 50% of sugar phosphates, 77% of carboxylic acids, and 72% of isomers, with low overall root mean square errors.

Conclusions:

• The developed mixed-mode chromatography method offers enhanced retention and separation for key metabolites.
• The MPNN model effectively predicts retention times, expanding the utility of metabolomics data analysis.
• This integrated approach aids in ranking metabolite identifications, particularly in complex biological samples like those from GLS2 knockout mouse primary hepatocytes.