Zachary Wu
@zacharywu.bsky.social
Researcher at Google DeepMind. Previously Arnold group at Caltech. Making useful molecules!
Reposted by Zachary Wu
Physics-based design of efficient Kemp eliminases
@lynnkamerlin.bsky.social
www.nature.com/articles/s41...
@lynnkamerlin.bsky.social
www.nature.com/articles/s41...
June 19, 2025 at 9:18 PM
Physics-based design of efficient Kemp eliminases
@lynnkamerlin.bsky.social
www.nature.com/articles/s41...
@lynnkamerlin.bsky.social
www.nature.com/articles/s41...
Reposted by Zachary Wu
The White House website has been wiped clean, but thanks to @erichorvitz.bsky.social , all the Biden #PCAST reports are available below. I want to thank the #PCAST members, all of whom donated their time and expertise to helping policy be science-based. tinyurl.com/2acnv2nh
PCAST Reports (2020-2025)
tinyurl.com
January 20, 2025 at 10:53 PM
The White House website has been wiped clean, but thanks to @erichorvitz.bsky.social , all the Biden #PCAST reports are available below. I want to thank the #PCAST members, all of whom donated their time and expertise to helping policy be science-based. tinyurl.com/2acnv2nh
Reposted by Zachary Wu
Reposted by Zachary Wu
A protein's position in the genome in relation to other genes is informative about function, so a genomic language model can be prompted to generate
- toxin/antitoxin pairs
- anti-CRISPR proteins
+ a database of 120B synthetic base pairs
www.biorxiv.org/content/10.1...
@brianhie.bsky.social
- toxin/antitoxin pairs
- anti-CRISPR proteins
+ a database of 120B synthetic base pairs
www.biorxiv.org/content/10.1...
@brianhie.bsky.social
December 18, 2024 at 10:14 PM
A protein's position in the genome in relation to other genes is informative about function, so a genomic language model can be prompted to generate
- toxin/antitoxin pairs
- anti-CRISPR proteins
+ a database of 120B synthetic base pairs
www.biorxiv.org/content/10.1...
@brianhie.bsky.social
- toxin/antitoxin pairs
- anti-CRISPR proteins
+ a database of 120B synthetic base pairs
www.biorxiv.org/content/10.1...
@brianhie.bsky.social
Reposted by Zachary Wu
Protein language models excel at generating functional yet remarkably diverse artificial sequences.
They however fail to naturally sample rare datapoints, like very high activities.
In our new preprint, we show that RL can solve this without the need for additional data:
arxiv.org/abs/2412.12979
They however fail to naturally sample rare datapoints, like very high activities.
In our new preprint, we show that RL can solve this without the need for additional data:
arxiv.org/abs/2412.12979
December 18, 2024 at 9:06 PM
Protein language models excel at generating functional yet remarkably diverse artificial sequences.
They however fail to naturally sample rare datapoints, like very high activities.
In our new preprint, we show that RL can solve this without the need for additional data:
arxiv.org/abs/2412.12979
They however fail to naturally sample rare datapoints, like very high activities.
In our new preprint, we show that RL can solve this without the need for additional data:
arxiv.org/abs/2412.12979
Reposted by Zachary Wu
Here's Veo 2, the latest version of our video generation model, as well as a substantial upgrade for Imagen 3 🧑🍳🚢
(Did I mention we are hiring on the Generative Media team, btw 👀)
blog.google/technology/g...
(Did I mention we are hiring on the Generative Media team, btw 👀)
blog.google/technology/g...
State-of-the-art video and image generation with Veo 2 and Imagen 3
We’re rolling out a new, state-of-the-art video model, Veo 2, and updates to Imagen 3. Plus, check out our new experiment, Whisk.
blog.google
December 16, 2024 at 5:35 PM
Here's Veo 2, the latest version of our video generation model, as well as a substantial upgrade for Imagen 3 🧑🍳🚢
(Did I mention we are hiring on the Generative Media team, btw 👀)
blog.google/technology/g...
(Did I mention we are hiring on the Generative Media team, btw 👀)
blog.google/technology/g...
Reposted by Zachary Wu
new preprint on chemical synthesis ML models
- showing how to combine multiple models in a principled way
- modern Transformers + GNN to featurize chemical reaction:
- new insights in where the models shine
+ bonus: find the quirky named reaction!
Feedback welcome!
arxiv.org/abs/2412.05269
- showing how to combine multiple models in a principled way
- modern Transformers + GNN to featurize chemical reaction:
- new insights in where the models shine
+ bonus: find the quirky named reaction!
Feedback welcome!
arxiv.org/abs/2412.05269
Chimera: Accurate retrosynthesis prediction by ensembling models with diverse inductive biases
Planning and conducting chemical syntheses remains a major bottleneck in the discovery of functional small molecules, and prevents fully leveraging generative AI for molecular inverse design. While ea...
arxiv.org
December 9, 2024 at 2:19 AM
new preprint on chemical synthesis ML models
- showing how to combine multiple models in a principled way
- modern Transformers + GNN to featurize chemical reaction:
- new insights in where the models shine
+ bonus: find the quirky named reaction!
Feedback welcome!
arxiv.org/abs/2412.05269
- showing how to combine multiple models in a principled way
- modern Transformers + GNN to featurize chemical reaction:
- new insights in where the models shine
+ bonus: find the quirky named reaction!
Feedback welcome!
arxiv.org/abs/2412.05269
Reposted by Zachary Wu
1/🧬 Excited to share PLAID, our new approach for co-generating sequence and all-atom protein structures by sampling from the latent space of ESMFold. This requires only sequences during training, which unlocks more data and annotations:
bit.ly/plaid-proteins
🧵
bit.ly/plaid-proteins
🧵
December 6, 2024 at 5:44 PM
1/🧬 Excited to share PLAID, our new approach for co-generating sequence and all-atom protein structures by sampling from the latent space of ESMFold. This requires only sequences during training, which unlocks more data and annotations:
bit.ly/plaid-proteins
🧵
bit.ly/plaid-proteins
🧵
Lots coming for protein design!
December 8, 2024 at 10:33 AM
Lots coming for protein design!
Reposted by Zachary Wu
Introducing Biomolecular Emulator (BioEmu). We build on AlphaFold evoformer representation of protein sequence, then sample 3D structures from equilibrium ensemble with diffusion model. From this sample, we can compute states, probabilities and understand mechanisms.
December 6, 2024 at 8:39 AM
Introducing Biomolecular Emulator (BioEmu). We build on AlphaFold evoformer representation of protein sequence, then sample 3D structures from equilibrium ensemble with diffusion model. From this sample, we can compute states, probabilities and understand mechanisms.
Reposted by Zachary Wu
Can a single small molecule rescue the stability of nearly all mutations in a protein?
Our new preprint by Taylor Mighell
www.biorxiv.org/content/10.1...
thread on the other place: x.com/taylor_mighe...
Our new preprint by Taylor Mighell
www.biorxiv.org/content/10.1...
thread on the other place: x.com/taylor_mighe...
A pharmacological chaperone stabilizer rescues the expression of the vast majority of pathogenic variants in a G protein-coupled receptor
Reduced protein stability is the most frequent mechanism by which rare missense variants cause disease. A promising therapeutic avenue for treating destabilizing variants is pharmacological chaperones...
www.biorxiv.org
December 3, 2024 at 9:40 AM
Can a single small molecule rescue the stability of nearly all mutations in a protein?
Our new preprint by Taylor Mighell
www.biorxiv.org/content/10.1...
thread on the other place: x.com/taylor_mighe...
Our new preprint by Taylor Mighell
www.biorxiv.org/content/10.1...
thread on the other place: x.com/taylor_mighe...
Reposted by Zachary Wu
This blog post by @drscotthawley.bsky.social provides a very accessible overview of flow matching / rectified flow and reflow, based on intuitions from physics, rather than starting from probability distributions. The visualisations and animations are excellent, and the whole thing is also a colab!
blog - Flow With What You Know
Basic physics provides a “straight, fast” way to get up to speed with flow-based generative models
drscotthawley.github.io
November 15, 2024 at 8:31 PM
This blog post by @drscotthawley.bsky.social provides a very accessible overview of flow matching / rectified flow and reflow, based on intuitions from physics, rather than starting from probability distributions. The visualisations and animations are excellent, and the whole thing is also a colab!
We also have an opening in Materials! Reminder: our London Wetlab position application closes Wednesday
We're hiring! Our team at Google DeepMind is looking for a research engineer to join us. More details in the link below.
boards.greenhouse.io/deepmind/job...
boards.greenhouse.io/deepmind/job...
Research Engineer, Materials Science
London, UK
boards.greenhouse.io
November 25, 2024 at 5:25 PM
We also have an opening in Materials! Reminder: our London Wetlab position application closes Wednesday
Reposted by Zachary Wu
Discovery of a biodegradable, fully recyclable plastic
www.science.org/doi/10.1126/...
www.science.org/doi/10.1126/...
Mechanically strong yet metabolizable supramolecular plastics by desalting upon phase separation
Plastics that can metabolize in oceans are highly sought for a sustainable future. In this work, we report the noncovalent synthesis of unprecedented plastics that are mechanically strong yet metaboli...
www.science.org
November 22, 2024 at 3:23 PM
Discovery of a biodegradable, fully recyclable plastic
www.science.org/doi/10.1126/...
www.science.org/doi/10.1126/...
Reposted by Zachary Wu
RFdiffusion for macrocycles www.biorxiv.org/content/10.1...
November 20, 2024 at 6:48 PM
RFdiffusion for macrocycles www.biorxiv.org/content/10.1...
Come join our London wet-lab team! Deadline to apply is next Wednesday boards.greenhouse.io/deepmind/job...
Laboratory Research Scientist, Protein Design, Science
London, UK
boards.greenhouse.io
November 18, 2024 at 3:47 PM
Come join our London wet-lab team! Deadline to apply is next Wednesday boards.greenhouse.io/deepmind/job...