Exploring the chemical subspace of RPLC: A data driven approach

Area of Science:

• Analytical Chemistry
• Environmental Science
• Computational Chemistry

Background:

• The human and environmental exposome contain numerous unknown chemical structures.
• Non-targeted analysis using liquid chromatography (LC) and high-resolution mass spectrometry is common for analyzing complex samples.
• The portion of chemical space retained and eluted by reversed-phase liquid chromatography (RPLC) is largely unknown.

Purpose of the Study:

• To develop a predictive model for assessing whether chemical compounds fall within the RPLC-amenable chemical space.
• To establish a data-driven framework for classifying molecules based on their RPLC retention behavior.

Main Methods:

• Construction of three retention index random forest (RF) regression models using molecular fingerprints to predict RPLC retention.
• Development of an RPLC RF classification model based on the regression models.
• Application of the classification model to a large dataset of small molecules (NORMAN SusDat).

Main Results:

• Molecular fingerprints effectively predict RPLC retention behavior.
• The RPLC classification model achieved 92% accuracy in predicting whether a chemical belongs to the RPLC subspace.
• 19.1% of 91,737 small molecules (≤1,000 Da) in the NORMAN SusDat dataset were predicted to fall 'outside' the RPLC subspace.

Conclusions:

• The developed RPLC chemical space model is a significant advancement in mapping chemical space.
• The model assesses the potential measurability of chemicals using RPLC methods.
• Identifying chemicals outside the RPLC subspace aids in reducing library searching candidates and avoids screening for unmeasurable compounds.