All-Atom Protein Generation with Latent Diffusion
Kaydedildi:
| Yayımlandı: | bioRxiv (Feb 13, 2025) |
|---|---|
| Yazar: | |
| Diğer Yazarlar: | , , , , , , , , |
| Baskı/Yayın Bilgisi: |
Cold Spring Harbor Laboratory Press
|
| Konular: | |
| Online Erişim: | Citation/Abstract Full Text - PDF Full text outside of ProQuest |
| Etiketler: |
Etiket eklenmemiş, İlk siz ekleyin!
|
MARC
| LEADER | 00000nab a2200000uu 4500 | ||
|---|---|---|---|
| 001 | 3141237419 | ||
| 003 | UK-CbPIL | ||
| 022 | |a 2692-8205 | ||
| 024 | 7 | |a 10.1101/2024.12.02.626353 |2 doi | |
| 035 | |a 3141237419 | ||
| 045 | 0 | |b d20250213 | |
| 100 | 1 | |a Lu, Amy X | |
| 245 | 1 | |a All-Atom Protein Generation with Latent Diffusion | |
| 260 | |b Cold Spring Harbor Laboratory Press |c Feb 13, 2025 | ||
| 513 | |a Working Paper | ||
| 520 | 3 | |a While generative models hold immense promise for protein design, existing models are typically backbone-only, despite the indispensable role that sidechain atoms play in mediating function. As prerequisite knowledge, all-atom 3D structure generation require the discrete sequence to specify sidechain identities, which poses a multimodal generation problem. We propose PLAID (Protein Latent Induced Diffusion), which samples from the latent space of a pre-trained sequence-tostructure predictor, ESMFold. The sampled latent embedding is then decoded with frozen decoders into the sequence and all-atom structure. Importantly, PLAID only requires sequence input during training, thus augmenting the dataset size by 2-4 orders of magnitude compared to the Protein Data Bank. It also makes more annotations available for functional control. As a demonstration of annotation-based prompting, we perform compositional conditioning on function and taxonomy using classifier-free guidance. Intriguingly, function-conditioned generations learn active site residue identities, despite them being non-adjacent on the sequence, and can correctly place the sidechains atoms. We further show that PLAID can generate transmembrane proteins with expected hydrophobicity patterns, perform motif scaffolding, and improve unconditional sample quality for long sequences. Links to model weights and training code are publicly available at github.com/amyxlu/plaid.Competing Interest StatementAXL, SAR, SK, VG, KC, RB, and NCF are employees of Genentech Inc., a member of the Roche Group.Footnotes* Title and formatting update: evaluation figures updated, motif scaffolding figure moved to main text, various other changes to content ordering | |
| 653 | |a Databases | ||
| 653 | |a Learning | ||
| 653 | |a Protein structure | ||
| 653 | |a Peptide mapping | ||
| 653 | |a Structure-function relationships | ||
| 653 | |a Axl protein | ||
| 653 | |a Amino acid sequence | ||
| 653 | |a Crystallography | ||
| 653 | |a Hydrophobicity | ||
| 653 | |a Proteins | ||
| 700 | 1 | |a Wilson, Yan | |
| 700 | 1 | |a Robinson, Sarah A | |
| 700 | 1 | |a Kelow, Simon | |
| 700 | 1 | |a Yang, Kevin K | |
| 700 | 1 | |a Gligorijevic, Vladimir | |
| 700 | 1 | |a Cho, Kyunghyun | |
| 700 | 1 | |a Bonneau, Richard | |
| 700 | 1 | |a Abbeel, Pieter | |
| 700 | 1 | |a Frey, Nathan C | |
| 773 | 0 | |t bioRxiv |g (Feb 13, 2025) | |
| 786 | 0 | |d ProQuest |t Biological Science Database | |
| 856 | 4 | 1 | |3 Citation/Abstract |u https://www.proquest.com/docview/3141237419/abstract/embedded/7BTGNMKEMPT1V9Z2?source=fedsrch |
| 856 | 4 | 0 | |3 Full Text - PDF |u https://www.proquest.com/docview/3141237419/fulltextPDF/embedded/7BTGNMKEMPT1V9Z2?source=fedsrch |
| 856 | 4 | 0 | |3 Full text outside of ProQuest |u https://www.biorxiv.org/content/10.1101/2024.12.02.626353v2 |