EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Training PBertKla on an Integrated Multi-Source Dataset with a Machine-Learning Layer for Lysine Lactylation Site Prediction

Seung Beom Jin, Junghee Park iD, Summer Dabin Lee, Ji Hye Han, Seung-Hyun Myung, Kichul Park iD, Jisoo Yun

International Journal of Molecular Sciences · 26.06.2026 · 27 (13) · 5761

DOI: 10.3390/ijms27135761 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Molecular Sciences, online erschienen am 26.06.2026. Verfasst von Seung Beom Jin und 6 weiteren Autor:innen. Beteiligte Einrichtung: LNPsolution, 199-9 Dugaebisan-ro, Hongcheon-gun 25114, Gangwon-do, Republic of Korea. Hauptthemen laut MEDLINE: Machine Learning, Lysine, Protein Processing, Post-Translational, Proteins. Die Arbeit nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Ministry of Education, Ministry of SMEs and Startups, Ministry of Science and ICT.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Lysine lactylation (Kla) is a recently discovered post-translational modification implicated in energy metabolism, cellular reprogramming, and disease progression. Here, we train the existing ProteinBERT-based predictor PBertKla on an integrated multi-source dataset and augment it with a lightweight machine-learning (ML) layer over sequence-derived features to predict Kla sites; on a common blind test set, the resulting model (PBertKla + ML) reaches an area under the receiver operating characteristic curve (AUROC) of 0.9126 on the integrated set and is statistically indistinguishable from the strongest available tool (Auto-Kla, DeLong p = 0.74) while significantly exceeding a recent ProtBert-based method (PCBert-Kla, p = 4 × 10−15). Two elements support this result. First, to train and benchmark the model, we assembled and released the largest curated Kla dataset to date, Multi (26,034 samples compiled from nine published sources through a 9-step quality-control pipeline), as a community resource. Second, we validated the model under a leakage-controlled protocol: re-training the complete pipeline under protein-level, 40%-identity homology, and leave-one-study-out splits—each verified to have zero train–test overlap—maintained ≈0.90 AUROC, only 0.6–1.5 percentage points (pp) below the random-split value, confirming genuine generalization rather than memorization. Ablation and SHapley Additive exPlanations (SHAP) analyses locate the predictive signal primarily in the ProteinBERT metafeature, with the ML layer adding a modest but real increment (+0.63 pp over PBertKla alone on Multi; no significant gain on the smaller hepatocellular carcinoma (HCC) set). Finally, an exploratory AlphaFold-based structural case study of FAM210A illustrates how predicted Kla sites distribute across ordered and disordered regions, without claiming a quantitative structure–probability relationship. All trained weights and code are publicly available.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Machine LearningLysineProtein Processing, Post-TranslationalProteinsComputational BiologyPredictive Learning ModelsPrediction AlgorithmsDatabases, ProteinHumans

Stichwörter der Autor:innen

AlphaFoldProteinBERTcurated benchmark datasetleakage-controlled validationlysine lactylationmachine learningpost-translational modification predictionprotein language model

Substanzen

LysineProteins

Autor:innen und Institutionen

  • LNPsolution, 199-9 Dugaebisan-ro, Hongcheon-gun 25114, Gangwon-do, Republic of Korea.
  • Department of Biochemistry and Molecular Biology, Chosun University School of Medicine, 309 Pilmoon-daero, Dong-gu, Gwangju 61452, Republic of Korea.
  • Institute of Well-Aging Medicare & Chosun University LAMP Center, Chosun University, Gwangju 61452, Republic of Korea.
  • Medical Research Institute & Chosun University LAMP Center, Chosun University, Gwangju 61452, Republic of Korea.

ORCID-Profile: Junghee Park, Kichul Park

Förderung

  • Ministry of Education
  • Ministry of SMEs and Startups
  • Ministry of Science and ICT
  • National Research Foundation of Korea
  • Government of the Republic of Korea
  • Korea Health Industry Development Institute

Zitieren

APA

Jin, S. B., Park, J., Lee, S. D., Han, J. H., Myung, S. H., Park, K., & Yun, J. (2026). Training PBertKla on an Integrated Multi-Source Dataset with a Machine-Learning Layer for Lysine Lactylation Site Prediction. International Journal of Molecular Sciences, 27(13), 5761. https://doi.org/10.3390/ijms27135761

Vancouver

Jin SB, Park J, Lee SD, Han JH, Myung SH, Park K, et al. Training PBertKla on an Integrated Multi-Source Dataset with a Machine-Learning Layer for Lysine Lactylation Site Prediction. International Journal of Molecular Sciences. 2026;27(13):5761. doi:10.3390/ijms27135761

BibTeX
@article{Jin2026,
  author = {Jin, Seung Beom and Park, Junghee and Lee, Summer Dabin and Han, Ji Hye and Myung, Seung-Hyun and Park, Kichul and Yun, Jisoo},
  title = {Training PBertKla on an Integrated Multi-Source Dataset with a Machine-Learning Layer for Lysine Lactylation Site Prediction},
  journal = {International Journal of Molecular Sciences},
  year = {2026},
  volume = {27},
  number = {13},
  pages = {5761},
  doi = {10.3390/ijms27135761}
}
Kostenlos starten →