Metabolite collision cross section prediction without energy-minimized structures.

Area of Science:

• Computational chemistry
• Analytical chemistry
• Metabolomics

Background:

• Ion mobility-mass spectrometry (IM-MS) is crucial for metabolite identification.
• Accurate collision cross section (CCS) values enhance confidence in IM-MS data interpretation.
• Computational prediction of CCS values can aid in matching experimental data.

Purpose of the Study:

• To develop accurate and accessible models for predicting metabolite collision cross section (CCS) values.
• To evaluate the performance of 2D vs. 3D molecular descriptors for CCS prediction.
• To enable CCS prediction without requiring energy-minimized structures.

Main Methods:

• Utilized linear methods for CCS prediction on the largest library of metabolite cross sections.
• Compared the efficacy of 2D and 3D molecular descriptors in predicting CCS.
• Validated predictions against experimental ion mobility-mass spectrometry data.

Main Results:

• Achieved a root mean square error of 7.0 Ų (median error of ~2%) for CCS predictions.
• Demonstrated the feasibility of accurate CCS prediction using simple, linear models.
• Showcased that 2D descriptors can be effective for CCS prediction, reducing the need for complex structural computations.

Conclusions:

• Simple linear models can accurately predict metabolite CCS values, improving metabolite identification in IM-MS.
• The study provides a valuable resource for CCS prediction, accessible to researchers.
• Predicting CCS using 2D descriptors offers a practical alternative to methods requiring energy-minimized structures.