EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Sparsity is all you need: rethinking biologically informed neural networks

Isabella Caranzano, Corrado Pancotti, Cesare Rollo, Flavio Sartori, Pietro Liò, Piero Fariselli, Tiziana Sanavia

Briefings in Bioinformatics · 03.07.2026 · 27 (4)

DOI: 10.1093/bib/bbag425 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Briefings in Bioinformatics, online erschienen am 06.08.2026. Verfasst von Isabella Caranzano und 6 weiteren Autor:innen. Beteiligte Einrichtung: AI and Computational Biomedicine Unit, Department of Medical Sciences, University of Torino, Via Santena 19, 10126 Torino, Italy. Hauptthemen laut MEDLINE: Neural Networks, Computer, Computational Biology. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 55 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0). Gefördert von: Italian Ministry of University and Research.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Biologically informed neural networks are increasingly adopted in bioinformatics under the premise that embedding biological knowledge into model architectures yields more accurate and interpretable predictions. This approach has driven a growing literature of pathway-informed models aiming to move beyond black-box learning by explicitly encoding biological structure. However, it remains unclear whether these models exploit biological knowledge or instead benefit from a different inductive bias. Here, we systematically investigate this question across 29 state-of-the-art pathway-informed neural networks by explicitly decoupling biological annotations from network architecture. For each evaluable model, we implement a structure-matched randomization protocol, in which pathway annotations are replaced with random associations while preserving sparsity and architectural constraints, allowing for a direct comparison under controlled conditions. Across multiple prediction tasks, datasets, and evaluation metrics, the randomized models consistently match or outperform their biologically informed counterparts. Moreover, pathway-informed models show no systematic advantage in interpretability: randomized models recover disease-associated biomarkers with comparable accuracy and yield highly correlated feature rankings. Our results reveal that the performance gains commonly attributed to biological pathway integration arise predominantly from sparsity-induced regularization rather than from biological knowledge itself. We provide a general evaluation workflow to test whether biological priors contribute predictive information beyond sparsity, offering practical guidance for the development of biology-aware neural networks. The code implementing the proposed methodology is available on GitHub at https://github.com/compbiomed-unito/Pathway_Randomization.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Neural Networks, ComputerComputational BiologyHumansAlgorithms

Stichwörter der Autor:innen

biologically informed neural networkscancer genomicsmodel randomization analysismulti-omics integrationpathway-based deep learning

Autor:innen und Institutionen

  • AI and Computational Biomedicine Unit, Department of Medical Sciences, University of Torino, Via Santena 19, 10126 Torino, Italy.
  • Helmholtz Zentrum München, Helmholtz AI Central Unit, Ingolstädter Landstraße 1, 85764 Oberschleißheim-Neuherberg, Germany.
  • Department of Computer Science, University of Copenhagen, Universitetsparken 1, 2100 Copenhagen, Denmark.
  • Center for Health Data Science, Department of Public Health, University of Copenhagen, Øster Søgade 16B, 10.1, 1355 Copenhagen, Denmark.
  • Department of Computer Science and Technology, University of Cambridge, William Gates Building, 15 JJ Thomson Ave, CB3 0FD Cambridge, United Kingdom.

Förderung

  • Italian Ministry of University and Research

Zitieren

APA

Caranzano, I., Pancotti, C., Rollo, C., Sartori, F., Liò, P., Fariselli, P., & Sanavia, T. (2026). Sparsity is all you need: rethinking biologically informed neural networks. Briefings in Bioinformatics, 27(4). https://doi.org/10.1093/bib/bbag425

Vancouver

Caranzano I, Pancotti C, Rollo C, Sartori F, Liò P, Fariselli P, et al. Sparsity is all you need: rethinking biologically informed neural networks. Briefings in Bioinformatics. 2026;27(4). doi:10.1093/bib/bbag425

BibTeX
@article{Caranzano2026,
  author = {Caranzano, Isabella and Pancotti, Corrado and Rollo, Cesare and Sartori, Flavio and Liò, Pietro and Fariselli, Piero and Sanavia, Tiziana},
  title = {Sparsity is all you need: rethinking biologically informed neural networks},
  journal = {Briefings in Bioinformatics},
  year = {2026},
  volume = {27},
  number = {4},
  doi = {10.1093/bib/bbag425}
}
Kostenlos starten →