Transfer learning in DeepLC improves LC retention time prediction across substantially different modifications and setups

TL;DR

This paper shows that transfer learning can improve the prediction of peptide retention times in liquid chromatography across different experimental setups and modifications.

Contribution

The novel use of transfer learning enables accurate LC retention time predictions even for substantially different modifications and setups.

Findings

Transfer learning significantly improves retention time prediction across diverse experimental conditions.

Fine-tuning pre-trained models adapts well to new peptide modifications and chromatography setups.

This approach enhances prediction accuracy for proteomics workflows with varying parameters.

Abstract

While LC retention time prediction of peptides and their modifications has proven useful, widespread adoption and optimal performance are hindered by variations in experimental parameters. These variations can render retention time prediction models inaccurate and dramatically reduce the value of predictions for identification, validation, and DIA spectral library generation. To date, mitigation of these issues has been attempted through calibration or by training bespoke models for specific experimental setups, with only partial success. We here demonstrate that transfer learning can successfully overcome these limitations by leveraging pre-trained model parameters. Remarkably, this approach can even fit highly performant models to substantially different peptide modifications and LC conditions than those on which the model was originally trained. This impressive adaptability of transfer learning makes it a highly robust solution for accurate peptide retention time prediction across a very wide variety of imaginable proteomics workflows.

LC retention time prediction of peptides and their modifications is useful but is hindered by variations in experimental parameters. Here, the authors show how fine-tuning a deep learning model on a wide variety of experimental setups and modified peptides substantially improves predictions.

Figures