Using Machine Learning to Predict Ion Mobility Collisional Cross Sections

Nowadays, ion mobility is getting more and more popular due to its ability to add an additional degree of separation when coupled with traditional techniques like liquid or gas chromatography along with providing key information for a molecule’s collision crosssection (CCS). CCS is a measurement of a molecule’s effective area when it is allowed to collide with a neutral gas such as Helium or Nitrogen under the influence of an electrical current. Uniquely, CCS values remain consistent regardless of sample preparation and analysis method allowing for it to be used as another layer to identify the molecule. ¹ One of the major limitations of using CCS to identify an unknown molecule is the lack of reference standards or values. As such, machine learning has become a promising alternative for the generation of predicted CCS values. To achieve this, two promising methods have been explored either using simulated geometric structures and the trajectory of the ion in the mobility cell, and more recently, building robust regression models using validated CCS values. Unfortunately, the utilization of geometric structures has been shown to be both slow and computationally expensive, with CCS deviations up to 30%. However, the utilization of robust regression models is a potential next step to solving this problem due to it reportedly fast calculation speed and comparatively high accuracy with an error up to 3%. ² 

References

1. Li, X.; Wang, H.; Jiang, M.; Ding, M.; Xu, X.; Xu, B.; Zou, Y.; Yu, Y.; Yang, W. Collision Cross Section Prediction Based on Machine Learning. Molecules 2023, 28 (10), 4050.
2. Song, X.-C.; Dreolin, N.; Canellas, E.; Goshawk, J.; Nerin, C. Prediction of Collision Cross-Section Values for Extractables and Leachables from Plastic Products. Environmental Science & Technology 2022, 56 (13), 9463-9473. DOI: 10.1021/acs.est.2c02853.