A Package to Predict Retention Times

The jp²rt package is a modern reimplementation of the Retip application described in

“Retip: Retention Time Prediction for Compound Annotation in Untargeted Metabolomics” by Paolo Bonini, Tobias Kind, Hiroshi Tsugawa, Dinesh Kumar Barupal, and Oliver Fiehn, Analytical Chemistry 2020 92 (11), 7515-7522 — 10.1021/acs.analchem.9b05765.

The basic idea of Retip is to train a machine learning regression model based on features that can be computed given the molecular structure (the molecular descriptors) on a reasonably small set of compounds for which the retention time is experimentally know, and then use such model to predict the retention time of a potentially very large set of new compounds.

With respect to the original implementation, the present one optimizes the computation of features, thanks to a robust native Java parallel implementation (see the API overview for more details) and replaces the somehow outdated R machine learning code (based on external dependencies that nowadays can be difficult to compile and run) with the modern Python state of the art machine learning stack. More in detail, this software has two components:

• the jp²rt Java package, allowing fast and reliable computation of molecular descriptors based on the Chemistry Development Kit,

• the jp²rt Python package, allowing model estimation and prediction, based on scikit-learn machine learning library.

As shown in the following documentation, both components can be used from the command line, or programmatically via the API that they expose.

This documentation consists of several key sections:

• Install: Instructions for software installation.

• A Worked Out Example: Guidance on selecting, estimating models, and using them for predicting retention times.

• Reference: Detailed information about jp²rt (including Java and Python APIs).

How to cite jp²rt

If you find this packge useful, please cite it as follows:

“jp²rt: A Java and Python package to Predict Retention Times” by Marianna Iorio^†, Sonia Maffioli^†, Massimo Santini^‡, and Matteo Simone^† — 10.5281/zenodo.10846234.

This work has been partially supported by Grant No. DM60066 from the Italian Ministry of University and Research to Naicons^† srl.