Structure-based drug design by denoising voxel grids

Pedro O. Pinheiro, Arian Rokkum Jamasb, Omar Mahmood, Vishnu Sresht, Saeed Saremi
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:40795-40812, 2024.

Abstract

We presents VoxBind, a new score-based generative model for 3D molecules conditioned on protein structures. Our approach represents molecules as 3D atomic density grids and leverages a 3D voxel-denoising network for learning and generation. We extend the neural empirical Bayes formalism (Saremi & Hyvärinen, 2019) to the conditional setting and generate structure-conditioned molecules with a two-step procedure: (i) sample noisy molecules from the Gaussian-smoothed conditional distribution with underdamped Langevin MCMC using the learned score function and (ii) estimate clean molecules from the noisy samples with single-step denoising. Compared to the current state of the art, our model is simpler to train, significantly faster to sample from, and achieves better results on extensive in silico benchmarks—the generated molecules are more diverse, exhibit fewer steric clashes, and bind with higher affinity to protein pockets.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-pinheiro24a, title = {Structure-based drug design by denoising voxel grids}, author = {Pinheiro, Pedro O. and Jamasb, Arian Rokkum and Mahmood, Omar and Sresht, Vishnu and Saremi, Saeed}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {40795--40812}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/pinheiro24a/pinheiro24a.pdf}, url = {https://proceedings.mlr.press/v235/pinheiro24a.html}, abstract = {We presents VoxBind, a new score-based generative model for 3D molecules conditioned on protein structures. Our approach represents molecules as 3D atomic density grids and leverages a 3D voxel-denoising network for learning and generation. We extend the neural empirical Bayes formalism (Saremi & Hyvärinen, 2019) to the conditional setting and generate structure-conditioned molecules with a two-step procedure: (i) sample noisy molecules from the Gaussian-smoothed conditional distribution with underdamped Langevin MCMC using the learned score function and (ii) estimate clean molecules from the noisy samples with single-step denoising. Compared to the current state of the art, our model is simpler to train, significantly faster to sample from, and achieves better results on extensive in silico benchmarks—the generated molecules are more diverse, exhibit fewer steric clashes, and bind with higher affinity to protein pockets.} }
Endnote
%0 Conference Paper %T Structure-based drug design by denoising voxel grids %A Pedro O. Pinheiro %A Arian Rokkum Jamasb %A Omar Mahmood %A Vishnu Sresht %A Saeed Saremi %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-pinheiro24a %I PMLR %P 40795--40812 %U https://proceedings.mlr.press/v235/pinheiro24a.html %V 235 %X We presents VoxBind, a new score-based generative model for 3D molecules conditioned on protein structures. Our approach represents molecules as 3D atomic density grids and leverages a 3D voxel-denoising network for learning and generation. We extend the neural empirical Bayes formalism (Saremi & Hyvärinen, 2019) to the conditional setting and generate structure-conditioned molecules with a two-step procedure: (i) sample noisy molecules from the Gaussian-smoothed conditional distribution with underdamped Langevin MCMC using the learned score function and (ii) estimate clean molecules from the noisy samples with single-step denoising. Compared to the current state of the art, our model is simpler to train, significantly faster to sample from, and achieves better results on extensive in silico benchmarks—the generated molecules are more diverse, exhibit fewer steric clashes, and bind with higher affinity to protein pockets.
APA
Pinheiro, P.O., Jamasb, A.R., Mahmood, O., Sresht, V. & Saremi, S.. (2024). Structure-based drug design by denoising voxel grids. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:40795-40812 Available from https://proceedings.mlr.press/v235/pinheiro24a.html.

Related Material