Packaged our algorithm to design binding peptides into a simple colab. Enter a protein sequence, indicate where you want the peptide to bind, and in a few minutes you'll have a peptide predicted to tightly bind via AlphaFold + Bayesian Optimization. 1/2

colab.research.google.com/github/ur-whit…

Just discovered Grant Sanderson videos and am now obsessed. Some examples: youtube.com/watch?v=pQa_tW… youtube.com/watch?v=X8jsij…

Join us in 1 week to hear Brian L Trippe and Jason Yim talk about diffusion models of protein backbones!

As always, sign up here for updates and zoom links: ml4proteinengineering.com

ML for protein design folks, what is the motivation behind encoding a continuous variable, such as the distance between two atoms, with a series of radial basis functions? It's a common encoding scheme (used in SCUBA, MPNN, etc), but these don't provide references/explanations

Diffusion models like #DALLE and #StableDiffusion are state of the art for image generation, yet our understanding of them is in its infancy. This thread introduces the basics of how diffusion models work, how we understand them, and why I think this understanding is broken.🧵

Awesome work! I made a Sergey Ovchinnikov 🇺🇦 -style Colab notebook for inpainting loops. Hope people find it useful! colab.research.google.com/github/polizzi…

Really cool sculpture of Top7 by Mike Tyka . Celebrating 10 years of IPD Institute for Protein Design

(check out the reflection in the glass) 🤯

Great and very thorough explanation of both theory and practice around diffusion-based generative models.
Slides: …2-tutorial-diffusion-models.github.io
Video: youtube.com/watch?v=cS6JQp…

Like a word processor’s autocomplete function, AI designed Institute for Protein Design draws on its understanding of how proteins fold to fill in additional parts of the protein around the central feature. bit.ly/3OEqfZc Jue Wang Ian Haydon

I ❤️ collaborative science!!! Our work on scaffolding chemically functional sites into de novo proteins by leveraging deep learning to jointly design protein sequence and structure is out! Let us know what you think.

What a joy it is to work with such great colleagues.

Super excited that this work is finally out. See Jue Wang twitter thread for a nice summary. If you need me I’ll be listening to “It’s plenty” by Burna Boy on repeat 🙌🏿👨🏿‍🔬👨🏿‍💻

Excited to see this published! 😎
PDF: ipd.uw.edu/wp-content/upl…
LINK: science.org/doi/10.1126/sc…
VIDEO: youtube.com/watch?v=-EJ8SX…
Also, see the thread by Jue Wang below:

One big leap between preprint and publication is that the inpainting method became very powerful and is now the method of choice for motif-scaffolding problems. Along with ProteinMPNN and AF2-based filtering, it has completely changed protein design at Institute for Protein Design. 8/8

Sidney Lisanza and I used hallucination to create protein binders, both by scaffolding a binding motif, and with only the target as input. This 'de novo' binder hallucination uses the ability of RosettaFold to model interactions between protein chains. 7/8

Doug Tischer linked together 2 binding sites to create a symmetric dimeric binder of the TrkA receptor involved in various diseases, showing that we can design binders to multiple targets in a geometrically defined way. This is crucial for modulating real signaling pathways. 6/8

We also successfully designed and tested metal- and protein-binding proteins. This shows the power of the inpainting method developed by Joseph Watson and David Juergens: from inputs with short helices or loops, we got outputs of well-packed, ideal helical bundles. 5/8