University of Alberta Scientists Develop New Online Server for Predicting LC–MS Retention Times

University of Alberta researchers developed a new web server for predicting the retention time (RT) of chemicals used in liquid chromatography–mass spectrometry (LC–MS) systems. Their findings were published in the Journal of Chromatography A (1).

Edmonton, Alberta - November 14, 2023: Buildings and scenery on the campus of the University of Alberta in Edmonton | Image Credit: © Torval Mork - stock.adobe.com

Liquid chromatography–mass spectrometry (LC–MS) is a highly effective analytical technique that excels at separating complex mixtures into individual components based on both chemical properties and mass-to-charge ratios (m/z). Nearly every LC–MS method uses high-performance (HPLC) or ultra-high-performance (UHPLC) systems in common modes such as reversed-phase liquid chromatography (RPLC) or hydrophilic interaction liquid chromatography (HILIC). By employing specially designed columns and stationary phases tailored to separate compounds through polarity-based interactions these systems can achieve fast, clean, and reproducible separations.

LC data can be used to provide chromatographic RT, which is a measure of the time it takes for a solute to pass through a chromatography column (2). RT can provide critical orthogonal information when aligned with structural data through prediction models. In RPLC and HILIC, RT is predominantly governed by analyte polarity; however, molecular size, shape, and flexibility can produce secondary effects depending on the stationary phase and analyte class. Combining RT with MS data can cause RT information to be used as a “molecular filter,” narrowing down the pool of potential compound matches, something that would be much larger if only MS data was utilized.

RT prediction was first explored with the 1980s, the scientists wrote. Developments in machine learning (ML) have improved the ability to predict RTs from known or postulated chemical structures, letting RT data be used more effectively in LC–MS-based compound identification. However, RT data is highly specific to chromatographic methods (CMs), with hundreds of different CMs with interdependent parameters being used; this limits the application of ML-based RT predictions in compound identification.

“Ideally, if RTs could be accurately predicted for a given CM and a given chemical structure (known or hypothesized), then the full potential of RTs in LC-MS analysis could be realized,” the scientists wrote (1).

In this study, a new RT prediction webserver, titled RT-Pred, was created to predict RTs for molecules across various chromatographic setups. According to the scientists, the system not only supports in-house CM-specific RT predictors, but it also enables users to train custom RT-Pred models using their own RT data on their own CMs, letting them predict RTs with their custom models. RT-Pred also supports RT and compound searches against its own database of millions of predicted RTs, spanning over 40 different CMs. RT-Pred also proved capable of accurately identifying compounds that either eluted in the void volume or was retained; if these classifiers were included, the system’s performance improved significantly.

The scientists expect to add more features to RT-Pred as time goes on, including expanding RT-Pred's CM coverage to incorporate a broader range of CMs, re-training some of the smaller CM models with more RT training examples, speeding up the training process using GPU accelerators, among other procedures. With these improvements, they aim to enable RT-Pred to support a wider variety of applications in industries such as metabolomics, foodomics, exposomics, natural product chemistry and analytical chemistry.

References

(1) Zakir, M.; LeVatte, M. A.; Wishart, D. S. RT-Pred: A Web Server for Accurate, Customized Liquid Chromatography Retention Time Prediction of Chemicals. J. Chromatogr. A 2025, 1747, 465816. DOI: 10.1016/j.chroma.2025.465816

(2) Understanding the Difference Between Retention Time and Relative Retention Time. ChromatographyToday 2025. https://www.chromatographytoday.com/news/autosamplers/36/breaking-news/understanding-the-difference-between-retention-time-and-relative-retention-time/31166 (accessed 2025-3-21)