EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

BaGGLS: a Bayesian shrinkage framework for interpretable modeling of interactions in high-dimensional biological data

Marta S Lemanczyk iD, Lucas Kock iD, Johanna Schlimme, Nadja Klein iD, Bernhard Y Renard iD

Bioinformatics · 2026

DOI: 10.1093/bioinformatics/btag441 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Bioinformatics, 2026. Verfasst von Marta S Lemanczyk und 4 weiteren Autor:innen. Beteiligte Einrichtung: Hasso Plattner Institute, Digital Engineering Faculty, University of Potsdam, Potsdam, Germany. Hauptthemen laut MEDLINE: Computational Biology, Software. Die Arbeit nennt 49 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: German Research Foundation, HPI Research School.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Motivation Biological data is often high dimensional, noisy, and governed by complex interactions among sparse signals. This poses major challenges for interpretability and reliable feature selection. Tasks such as identifying motif interactions in genomics exemplify these difficulties, as only a small subset of biologically relevant features (e.g. motifs) are typically active, and their effects are often non-linear and context-dependent. While statistical approaches often result in more interpretable models, deep learning models have proven effective in modeling complex interactions and prediction accuracy, yet their black-box nature limits interpretability. Results We introduce BaGGLS, a flexible and interpretable probabilistic binary regression model designed for high-dimensional biological inference involving feature interactions. BaGGLS incorporates a Bayesian group global-local shrinkage prior, aligned with the group structure introduced by interaction terms. This prior encourages sparsity while retaining interpretability, helping to isolate meaningful signals and suppress noise. To enable scalable inference, we employ a partially factorized variational approximation that captures posterior skewness and supports efficient learning even in large feature spaces. In extensive simulations, we compare BaGGLS to frequentist probit regressions (unconstrained and with L1-penalty) as well as a probit model with Markov Chain Monte Carlo (MCMC) sampling under a horseshoe prior. We can show that BaGGLS outperforms the other methods with regard to interaction detection and is many times faster than MCMC sampling under the horseshoe prior. We also demonstrate the usefulness of BaGGLS in the context of interaction discovery from motif scanner outputs (e.g. Find Individual Motif Occurrences (FIMO)) and noisy attribution scores from deep learning models. This shows that BaGGLS is a promising approach for uncovering biologically relevant interaction patterns, with potential applicability across a range of high-dimensional tasks in computational biology. Availability Code is available at gitlab.com/dacs-hpi/baggls.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Computational BiologySoftwareBayes TheoremAlgorithmsDeep Learning

Autor:innen und Institutionen

  • Hasso Plattner Institute, Digital Engineering Faculty, University of Potsdam, Potsdam, Germany.
  • Department of Statistics and Data Science, National University of Singapore, Singapore, Singapore.
  • Scientific Computing Center, Karlsruhe Institute of Technology, Karlsruhe, Germany.

ORCID-Profile: Marta S Lemanczyk, Lucas Kock, Nadja Klein, Bernhard Y Renard

Förderung

  • German Research Foundation
  • HPI Research School

Zitieren

APA

Lemanczyk, M. S., Kock, L., Schlimme, J., Klein, N., & Renard, B. Y. (2026). BaGGLS: a Bayesian shrinkage framework for interpretable modeling of interactions in high-dimensional biological data. Bioinformatics. https://doi.org/10.1093/bioinformatics/btag441

Vancouver

Lemanczyk MS, Kock L, Schlimme J, Klein N, Renard BY. BaGGLS: a Bayesian shrinkage framework for interpretable modeling of interactions in high-dimensional biological data. Bioinformatics. 2026. doi:10.1093/bioinformatics/btag441

BibTeX
@article{Lemanczyk2026,
  author = {Lemanczyk, Marta S and Kock, Lucas and Schlimme, Johanna and Klein, Nadja and Renard, Bernhard Y},
  title = {BaGGLS: a Bayesian shrinkage framework for interpretable modeling of interactions in high-dimensional biological data},
  journal = {Bioinformatics},
  year = {2026},
  doi = {10.1093/bioinformatics/btag441}
}
Kostenlos starten →