EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Impact of training data on deep learning–based RNA 3D structure quality assessment

Bartosz Adamczyk iD, Marta Szachniuk iD, Maciej Antczak iD

Briefings in Bioinformatics · 02.11.2026 · 27 (6)

DOI: 10.1093/bib/bbag630 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Briefings in Bioinformatics, online erschienen am 09.11.2026. Verfasst von Bartosz Adamczyk und 2 weiteren Autor:innen. Beteiligte Einrichtung: Institute of Bioorganic Chemistry, Polish Academy of Sciences, Noskowskiego 12/14, PL-61-704 Poznan, Poland. Hauptthemen laut MEDLINE: Deep Learning, RNA, Nucleic Acid Conformation, Computational Biology. Die Arbeit nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Science Centre, Institute of Computing Science, Poznan University of Technology.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate reference-free assessment of predicted RNA 3D structures is essential for large-scale RNA modeling, yet the impact of training data on deep learning–based quality predictors remains poorly understood. Here, we present a systematic cross-training analysis of three deep learning predictors for RNA 3D structure quality assessment, each originally developed on distinct training sets. By retraining each predictor on alternative (nonnative) datasets—including RNA3DQA, a new set created specifically for this project—and evaluating performance on independent RNA-Puzzles benchmarks, we systematically assess the relative contributions of model architecture and training data composition. Our results reveal pronounced differences in cross-dataset generalization, with predictors often performing better when trained on nonnative datasets than on their native ones. Notably, the impact of the training datasets varied across predictors, with RNA3DQA yielding the overall best accuracy and ranking stability results specifically for the ARES model. Native training sets yielded better performance for RNA3DCNN and lociPARSE. These findings demonstrate that the characteristics of the training data play a dominant, model-dependent role in shaping the performance and robustness of reference-free RNA 3D structure quality assessment methods. We conclude with practical recommendations for designing transferable training datasets and for benchmarking deep learning–based quality predictors in RNA structural bioinformatics.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Deep LearningRNANucleic Acid ConformationComputational BiologyModels, Molecular

Stichwörter der Autor:innen

RNA 3D structure predictiondeep learninggeneral quality assessmentreference-free evaluation

Substanzen

RNA

Autor:innen und Institutionen

  • Institute of Bioorganic Chemistry, Polish Academy of Sciences, Noskowskiego 12/14, PL-61-704 Poznan, Poland.
  • Institute of Computing Science, Poznan University of Technology, Jacka Rychlewskiego 1, PL-61-131 Poznan, Poland.

ORCID-Profile: Bartosz Adamczyk, Marta Szachniuk, Maciej Antczak

Förderung

  • National Science Centre
  • Institute of Computing Science, Poznan University of Technology

Zitieren

APA

Adamczyk, B., Szachniuk, M., & Antczak, M. (2026). Impact of training data on deep learning–based RNA 3D structure quality assessment. Briefings in Bioinformatics, 27(6). https://doi.org/10.1093/bib/bbag630

Vancouver

Adamczyk B, Szachniuk M, Antczak M. Impact of training data on deep learning–based RNA 3D structure quality assessment. Briefings in Bioinformatics. 2026;27(6). doi:10.1093/bib/bbag630

BibTeX
@article{Adamczyk2026,
  author = {Adamczyk, Bartosz and Szachniuk, Marta and Antczak, Maciej},
  title = {Impact of training data on deep learning–based RNA 3D structure quality assessment},
  journal = {Briefings in Bioinformatics},
  year = {2026},
  volume = {27},
  number = {6},
  doi = {10.1093/bib/bbag630}
}
Kostenlos starten →