Development of a Reverse Phase HPLC Retention Index Model for Nontargeted Metabolomics Using Synthetic Compounds.

Area of Science:

• Analytical Chemistry
• Chemometrics
• Metabolomics

Background:

• Non-targeted metabolomics using High-Performance Liquid Chromatography (HPLC) requires robust tools for structure identification.
• Existing methods for predicting metabolite retention behavior can be limited in accuracy and scope.

Purpose of the Study:

• To develop a more accurate HPLC retention index model for the MolFind application.
• To improve the prediction of human metabolite structures using synthetic screening compounds.

Main Methods:

• Development of an artificial neural network (ANN) model trained on 1955 measured synthetic compounds.
• Independent validation using a set of 202 human metabolites.
• Introduction of a Qualitative Range of Interest (QRI) modification for neural network training.

Main Results:

• The new ANN model demonstrated improved accuracy compared to previous models.
• The model successfully predicted complex human metabolites not similar to the training data.
• High sensitivity achieved: 97% for metabolites with feature combinations in >=3 training compounds, >90% for others.

Conclusions:

• Inexpensive synthetic compounds can effectively train accurate predictive models for metabolomics.
• The developed model enhances structure identification capabilities in non-targeted metabolomics.
• The QRI modification allows simultaneous quantitative and qualitative data integration in predictive modeling.