Flexible Kernels for Protein Property Prediction

Martin Jankowiak, Yerdos Ordabayev, Rudraksh Tuwani, Henry Neil Ward, Hunter Nisonoff, James M Mcfarland, Gevorg Grigoryan
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:50779-50829, 2026.

Abstract

Despite its importance to applications in protein design, predicting protein properties like binding affinity and thermostability from sparse experimental data remains a significant challenge. Accordingly, we introduce a class of sequence kernels that exploit evolutionary substitution matrices as well as local linearity and demonstrate that the resulting Gaussian processes provide data-efficient models of protein property landscapes, frequently outperforming alternatives that rely on foundation model embeddings. Furthermore–by learning what are in effect structure-aware substitution matrices–we show that our kernels can readily incorporate structural information from foundation models. We demonstrate that these structure-conditioned kernels are well suited to multi-task learning across multiple protein property landscapes and can decisively outperform local supervised learning methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-jankowiak26a, title = {Flexible Kernels for Protein Property Prediction}, author = {Jankowiak, Martin and Ordabayev, Yerdos and Tuwani, Rudraksh and Ward, Henry Neil and Nisonoff, Hunter and Mcfarland, James M and Grigoryan, Gevorg}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {50779--50829}, 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/jankowiak26a/jankowiak26a.pdf}, url = {https://proceedings.mlr.press/v306/jankowiak26a.html}, abstract = {Despite its importance to applications in protein design, predicting protein properties like binding affinity and thermostability from sparse experimental data remains a significant challenge. Accordingly, we introduce a class of sequence kernels that exploit evolutionary substitution matrices as well as local linearity and demonstrate that the resulting Gaussian processes provide data-efficient models of protein property landscapes, frequently outperforming alternatives that rely on foundation model embeddings. Furthermore–by learning what are in effect structure-aware substitution matrices–we show that our kernels can readily incorporate structural information from foundation models. We demonstrate that these structure-conditioned kernels are well suited to multi-task learning across multiple protein property landscapes and can decisively outperform local supervised learning methods.} }
Endnote
%0 Conference Paper %T Flexible Kernels for Protein Property Prediction %A Martin Jankowiak %A Yerdos Ordabayev %A Rudraksh Tuwani %A Henry Neil Ward %A Hunter Nisonoff %A James M Mcfarland %A Gevorg Grigoryan %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-jankowiak26a %I PMLR %P 50779--50829 %U https://proceedings.mlr.press/v306/jankowiak26a.html %V 306 %X Despite its importance to applications in protein design, predicting protein properties like binding affinity and thermostability from sparse experimental data remains a significant challenge. Accordingly, we introduce a class of sequence kernels that exploit evolutionary substitution matrices as well as local linearity and demonstrate that the resulting Gaussian processes provide data-efficient models of protein property landscapes, frequently outperforming alternatives that rely on foundation model embeddings. Furthermore–by learning what are in effect structure-aware substitution matrices–we show that our kernels can readily incorporate structural information from foundation models. We demonstrate that these structure-conditioned kernels are well suited to multi-task learning across multiple protein property landscapes and can decisively outperform local supervised learning methods.
APA
Jankowiak, M., Ordabayev, Y., Tuwani, R., Ward, H.N., Nisonoff, H., Mcfarland, J.M. & Grigoryan, G.. (2026). Flexible Kernels for Protein Property Prediction. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:50779-50829 Available from https://proceedings.mlr.press/v306/jankowiak26a.html.

Related Material