Objective:

To develop and evaluate a method for predicting liquid chromatography retention times across different experimental systems.

Approach:

• Two-step prediction: The method first predicts the order in which compounds leave the column using a graph neural network informed by molecular structure, column information, and mobile-phase pH. It then maps that order to absolute retention times using a polynomial calibration based on confidently identified compounds from the target dataset.
• Calibration and evaluation: Calibration used approximately 30 confidently identified compounds as anchors; the researchers found that as few as 15 could be sufficient. The method was trained and tested on 171 curated reversed-phase chromatography datasets from RepoRT, with testing conditions separated from those used for training.
• Application: The researchers applied the method to liquid chromatography-tandem mass spectrometry data from human fecal samples, using 19 known compounds as anchors.

Key Findings:

• The condition-aware model’s retention-order error rate was 4.89%, compared with 5.61% for a similar model that did not account for chromatographic conditions.
• When structurally similar compounds were separated between training and test sets, the two-step method outperformed two existing models in five of six datasets and performed similarly in the sixth; it did not require training data from the target system.
• In fecal-sample data, retention-time predictions provided additional evidence for compound annotations, helped evaluate candidate bile acids, and supported differentiation of stereoisomers.
• Column chemistry had the largest effect on retention order. Mobile-phase pH, solvent composition, additives, and temperature also affected some compounds.

Interpretation:

The method predicts compound elution order before calibrating that order to a laboratory’s measured retention times. The researchers report that experimental conditions, particularly column chemistry, influence retention behavior.

Limitations:

• The method is currently limited to reversed-phase chromatography.
• It depends on accurate experimental metadata and reliable anchor identifications.

Conclusion:

The researchers note the importance of recording column, mobile phase, gradient, temperature, and flow rate when generating or sharing retention data.

Sources:

• Article context; study published in Nature Methods