(Be Cautious!) Bio-Foundation Models Are Not Yet Robust to Biologically Plausible Perturbations and ML Transformations

Jinhao Duan, Ruichen Zhang, Gengwei Zhang, Huaizhi Qu, Jie Peng, Sijia Liu, Tianlong Chen
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:27045-27068, 2026.

Abstract

Though biological foundation models (Bio-FMs) have delivered strong performance across biomedical tasks, their robustness to small-but-real perturbations is underexplored. In this work, we ask: Are Bio-FMs robust for real-world use? What perturbations compromise their reliability? Our pilot study suggests that due to subtle biological data curation issues and common machine-learning (ML) processing choices, Bio-FMs suffer from two complementary perturbation sources: biologically plausible perturbations (capturing experimental corruptions and curation artifacts) and ML-induced transformations (capturing preprocessing, data augmentation, and embedding choices). Guided by this taxonomy, we design perturbation suites that mimic corruptions frequently encountered in biological experiments, and we systematically probe how transformations in the ML pipeline reshape model behavior. By conducting 2,128 experiments over 11 state-of-the-art Bio-FMs on 7 bio-tasks, we show that most Bio-FMs are vulnerable to both biological perturbations and ML transformations, revealing underappreciated robustness gaps that can directly translate into deployment risk. Interestingly, we find that subtle biological perturbations, which are often imperceptible to current measurement tools, can induce severe discrepancies in Bio-FM outputs and lead to critical failures, yet cryo-EM models (e.g., CryoDRGN) exhibit a surprising level of robustness even under worst-case perturbations. Our study for the first time surfaces critical failure modes and provides a principled perspective for evaluating the robustness of Bio-FMs.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-duan26f, title = {({B}e Cautious!) Bio-Foundation Models Are Not Yet Robust to Biologically Plausible Perturbations and {ML} Transformations}, author = {Duan, Jinhao and Zhang, Ruichen and Zhang, Gengwei and Qu, Huaizhi and Peng, Jie and Liu, Sijia and Chen, Tianlong}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {27045--27068}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/duan26f/duan26f.pdf}, url = {https://proceedings.mlr.press/v306/duan26f.html}, abstract = {Though biological foundation models (Bio-FMs) have delivered strong performance across biomedical tasks, their robustness to small-but-real perturbations is underexplored. In this work, we ask: Are Bio-FMs robust for real-world use? What perturbations compromise their reliability? Our pilot study suggests that due to subtle biological data curation issues and common machine-learning (ML) processing choices, Bio-FMs suffer from two complementary perturbation sources: biologically plausible perturbations (capturing experimental corruptions and curation artifacts) and ML-induced transformations (capturing preprocessing, data augmentation, and embedding choices). Guided by this taxonomy, we design perturbation suites that mimic corruptions frequently encountered in biological experiments, and we systematically probe how transformations in the ML pipeline reshape model behavior. By conducting 2,128 experiments over 11 state-of-the-art Bio-FMs on 7 bio-tasks, we show that most Bio-FMs are vulnerable to both biological perturbations and ML transformations, revealing underappreciated robustness gaps that can directly translate into deployment risk. Interestingly, we find that subtle biological perturbations, which are often imperceptible to current measurement tools, can induce severe discrepancies in Bio-FM outputs and lead to critical failures, yet cryo-EM models (e.g., CryoDRGN) exhibit a surprising level of robustness even under worst-case perturbations. Our study for the first time surfaces critical failure modes and provides a principled perspective for evaluating the robustness of Bio-FMs.} }
Endnote
%0 Conference Paper %T (Be Cautious!) Bio-Foundation Models Are Not Yet Robust to Biologically Plausible Perturbations and ML Transformations %A Jinhao Duan %A Ruichen Zhang %A Gengwei Zhang %A Huaizhi Qu %A Jie Peng %A Sijia Liu %A Tianlong Chen %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-duan26f %I PMLR %P 27045--27068 %U https://proceedings.mlr.press/v306/duan26f.html %V 306 %X Though biological foundation models (Bio-FMs) have delivered strong performance across biomedical tasks, their robustness to small-but-real perturbations is underexplored. In this work, we ask: Are Bio-FMs robust for real-world use? What perturbations compromise their reliability? Our pilot study suggests that due to subtle biological data curation issues and common machine-learning (ML) processing choices, Bio-FMs suffer from two complementary perturbation sources: biologically plausible perturbations (capturing experimental corruptions and curation artifacts) and ML-induced transformations (capturing preprocessing, data augmentation, and embedding choices). Guided by this taxonomy, we design perturbation suites that mimic corruptions frequently encountered in biological experiments, and we systematically probe how transformations in the ML pipeline reshape model behavior. By conducting 2,128 experiments over 11 state-of-the-art Bio-FMs on 7 bio-tasks, we show that most Bio-FMs are vulnerable to both biological perturbations and ML transformations, revealing underappreciated robustness gaps that can directly translate into deployment risk. Interestingly, we find that subtle biological perturbations, which are often imperceptible to current measurement tools, can induce severe discrepancies in Bio-FM outputs and lead to critical failures, yet cryo-EM models (e.g., CryoDRGN) exhibit a surprising level of robustness even under worst-case perturbations. Our study for the first time surfaces critical failure modes and provides a principled perspective for evaluating the robustness of Bio-FMs.
APA
Duan, J., Zhang, R., Zhang, G., Qu, H., Peng, J., Liu, S. & Chen, T.. (2026). (Be Cautious!) Bio-Foundation Models Are Not Yet Robust to Biologically Plausible Perturbations and ML Transformations. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:27045-27068 Available from https://proceedings.mlr.press/v306/duan26f.html.

Related Material