A High-Quality Acetylation Dataset Reveals Modest Data Requirements for Transfer Learning to Identify Little Studied Post-Translational Modifications
- 1Hasso Plattner Institute
Abstract
Dysregulation of post-translational modifications (PTMs) is associated with severe pathologies, including cancers and Alzheimer's disease. Despite their biological importance, identifying modified peptides remains challenging due to the immense combinatorial search space. While searches benefit from prior knowledge of a peptide's modification status, the data scarcity for most PTMs hinders the development of accurate deep learning classifiers like AHLF (ad hoc learning of peptide fragmentation). Here, we overcome this data bottleneck for acetylation and ubiquitination. We harmonised a dataset with about 500,000 high quality acetylated peptide-spectrum matches (PSMs) from nine publicly available acetylation-enriched datasets. We fine-tuned AHLF with the acetylation and a 2-million spectra strong ubiquitination dataset separately and assessed the minimum data requirement for training by iteratively downsampling. Training separate models on SILAC and label-free subsets also assessed the impact of data diversity. The resulting acetylation and ubiquitination models achieve an AUC of 0.87 and 0.90 respectively. Beyond 28,500 acetylated spectra, corresponding to roughly 0.3% of the original model's training data, additional data just provides minor performance gains. Finally, we show that data diversity is beneficial for generalizability, while models trained on homogeneous data sources tend to overfit to their respective data type. All code, and model weights are available at https://gitlab.com/dacs-hpi/ahlf-ptmai.
Related
- Oxidative Stress, Post‐Translational Modifications, and Type 1 Diabetes Pathogenesis
- Post-translational modifications in rosacea
- A delayed translocation into the endoplasmic reticulum controls the post-translational modifications of PD-L1
- Post-translational modifications of PLK1: a dynamic regulatory network in cell cycle regulation and tumor progression
- ProteomeScout v4: An open-source, harmonized, ecosystem for post-translational modifications
BibTeX
@misc{luplow2026highquality,
title = {{A High-Quality Acetylation Dataset Reveals Modest Data Requirements for Transfer Learning to Identify Little Studied Post-Translational Modifications}},
author = {Y. Luplow and S. Wang and A. R. Jones and J. A. Vizcaino and C. N. Schlaffner and B. Y. Renard},
howpublished = {bioRxiv},
year = {2026},
month = jun,
doi = {10.64898/2026.06.25.733353},
url = {https://www.biorxiv.org/content/10.64898/2026.06.25.733353v3},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - A High-Quality Acetylation Dataset Reveals Modest Data Requirements for Transfer Learning to Identify Little Studied Post-Translational Modifications
AU - Y. Luplow
AU - S. Wang
AU - A. R. Jones
AU - J. A. Vizcaino
AU - C. N. Schlaffner
AU - B. Y. Renard
PY - 2026
DA - 2026/06/25
PB - bioRxiv
DO - 10.64898/2026.06.25.733353
UR - https://www.biorxiv.org/content/10.64898/2026.06.25.733353v3
AB - Dysregulation of post-translational modifications (PTMs) is associated with severe pathologies, including cancers and Alzheimer's disease. Despite their biological importance, identifying modified peptides remains challenging due to the immense combinatorial search space. While searches benefit from prior knowledge of a peptide's modification status, the data scarcity for most PTMs hinders the development of accurate deep learning classifiers like AHLF (ad hoc learning of peptide fragmentation). Here, we overcome this data bottleneck for acetylation and ubiquitination. We harmonised a dataset with about 500,000 high quality acetylated peptide-spectrum matches (PSMs) from nine publicly available acetylation-enriched datasets. We fine-tuned AHLF with the acetylation and a 2-million spectra strong ubiquitination dataset separately and assessed the minimum data requirement for training by iteratively downsampling. Training separate models on SILAC and label-free subsets also assessed the impact of data diversity. The resulting acetylation and ubiquitination models achieve an AUC of 0.87 and 0.90 respectively. Beyond 28,500 acetylated spectra, corresponding to roughly 0.3% of the original model's training data, additional data just provides minor performance gains. Finally, we show that data diversity is beneficial for generalizability, while models trained on homogeneous data sources tend to overfit to their respective data type. All code, and model weights are available at https://gitlab.com/dacs-hpi/ahlf-ptmai.
ER -