Cross-domain transfer learning from peptides to metabolites using a multi-property fine-tuned LLM
Bioinformatics · 02.07.2026 · 42 (7)
DOI: 10.1093/bioinformatics/btag493 ↗Kurz gefasst
Fachartikel in der Zeitschrift Bioinformatics, online erschienen am 03.07.2026. Verfasst von Uchenna Alex Anyaegbunam und 6 weiteren Autor:innen. Beteiligte Einrichtung: Computational Biology and Data Mining group (CBDM), Institute of Organismic and Molecular Evolution (iOME), Johannes Gutenberg University, Mainz, 551…. Hauptthemen laut MEDLINE: Peptides, Metabolomics. Die Arbeit nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Federal Ministry for Education and Research (BMBF): curATime, diAMs, DIASyM.
Abstract (Originalsprache)
Motivation Accurate liquid chromatography retention time (RT) prediction is a critical component of compound identification in metabolomics and lipidomics. However, existing RT prediction approaches are often limited by the scarcity of experimental RT measurements for many molecular classes, restricting model generalization and the construction of comprehensive RT libraries. Transfer learning from data-rich chemical domains offers a potential strategy to overcome these limitations, but its effectiveness for metabolite RT prediction remains insufficiently explored. Results We developed a transfer learning framework based on ChemBERTa that leverages large peptide datasets to improve metabolite RT prediction under data-sparse conditions. A peptide-pretrained model was trained using a multi-task objective that jointly predicted RT and seven RDKit-derived molecular descriptors. Compared with an RT-only model, the multi-task approach learned more robust chemical representations and demonstrated superior generalization to metabolites, achieving a median test R² of 0.842 versus 0.820. When transferred to metabolite RT prediction, the multi-task pretrained model substantially outperformed models trained from scratch at low-data regimes. Using only 3% of metabolite training data (2129 compounds), transfer learning achieved a median test R² of 0.322 compared with 0.216 for the baseline model, while reducing MAE from 131.7 to 114.9. Significant improvements were also observed at 5% and 10% training fractions, with benefits gradually diminishing as larger metabolite datasets became available. In contrast, a peptide-pretrained single-task RT model showed performance comparable to the baseline, indicating that the observed gains arise primarily from multi-task molecular property learning rather than peptide pretraining alone. These findings demonstrate that multi-task transfer learning provides an effective and scalable strategy for improving RT prediction in metabolomics, particularly when experimental training data are limited. Availability Freely available on https://github.com/uchealex/CHEMBEDDING.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Substanzen
Autor:innen und Institutionen
- Computational Biology and Data Mining group (CBDM), Institute of Organismic and Molecular Evolution (iOME), Johannes Gutenberg University, Mainz, 55128, Germany.
- Institute of Computer Science, Johannes-Gutenberg University, Mainz, 55128, Germany.
- Institute for Quantitative and Computational Biosciences (IQCB), Johannes-Gutenberg University, Mainz, 55128, Germany.
- Institute of Immunology, University Medical Center of the Johannes-Gutenberg University Mainz, Mainz, 55131, Germany.
- Research Centre for Immunotherapy (FZI), University Medical Center of the Johannes-Gutenberg University Mainz, Mainz, 55131, Germany.
- Department of Dermatology, University Medical Center of the Johannes Gutenberg University Mainz, Mainz, 55131, Germany.
ORCID-Profile: Uchenna Alex Anyaegbunam, David Teschner, Maximilian Sprang, Miguel A Andrade-Navarro
Förderung
- Federal Ministry for Education and Research (BMBF): curATime
- diAMs
- DIASyM
- Forschungskernen für Massenspektrometrie in der Systemmedizin (MSCoreSys)
- ReALity Initiative of the Johannes Gutenberg Universität Mainz and the Forschungsinitiative des Landes Rheinland-Pfalz
- Einstein Early Career Researcher
Zitieren
APA
Anyaegbunam, U. A., Teschner, D., Schmidlin, T., Hildebrandt, A., Mayer, J. U., Sprang, M., & Andrade-Navarro, M. A. (2026). Cross-domain transfer learning from peptides to metabolites using a multi-property fine-tuned LLM. Bioinformatics, 42(7). https://doi.org/10.1093/bioinformatics/btag493
Vancouver
Anyaegbunam UA, Teschner D, Schmidlin T, Hildebrandt A, Mayer JU, Sprang M, et al. Cross-domain transfer learning from peptides to metabolites using a multi-property fine-tuned LLM. Bioinformatics. 2026;42(7). doi:10.1093/bioinformatics/btag493
BibTeX
@article{Anyaegbunam2026,
author = {Anyaegbunam, Uchenna Alex and Teschner, David and Schmidlin, Thierry and Hildebrandt, Andreas and Mayer, Johannes U and Sprang, Maximilian and Andrade-Navarro, Miguel A},
title = {Cross-domain transfer learning from peptides to metabolites using a multi-property fine-tuned LLM},
journal = {Bioinformatics},
year = {2026},
volume = {42},
number = {7},
doi = {10.1093/bioinformatics/btag493}
}
EUVIMED