Gaurav Bhardwaj

RFpeptides made it to the cover of Nature Chemical Biology December issue!

Credit to Stephen Rettie for leading the work and making this very cool graphic!

Cover image of the December issue. It depicts a diffusion-based generative deep learning pipeline for de novo design of a macrocylic peptide targeting the protein, myeloid cell leukemia 1.

Our December issue is live!

www.nature.com/nchembio/vol...

The cover depicts a diffusion-based generative deep learning pipeline for de novo design of a macrocyclic peptide targeting the protein, myeloid cell leukemia 1

Glad that we could contribute to this very cool work by @yehlincho.bsky.social and @sokrypton.org! See the nice skeetorial below by Yehlin. More to come soon with experimental validation and exciting applications in protein and peptide design.

Accelerating Biomolecular Modeling with AtomWorks and RF3 Deep learning methods trained on protein structure databases have revolutionized biomolecular structure prediction, but developing and training new models remains a considerable challenge. To facilita...

(1/7)
Training biomolecular foundation models shouldn't be so hard. And open-source structure prediction is important. So today we're releasing two software packages: AtomWorks and RosettaFold3 (RF3)

[https://www.biorxiv.org/content/10.1101/2025.08.14.670328v2](www.biorxiv.org/content/10.1...)

We are very excited to announce that early bird registration for European RosettaCon 2025 is now open!

More information here: europeanrosettacon.org

Thank you!! Look forward to meeting you in Grenoble soon.

Cyclic peptide structure prediction and design using AlphaFold2 Nature Communications - AfCycDesign: Cyclic offset to the relative positional encoding in AlphaFold2 enables accurate structure prediction, sequence redesign, and de novo hallucination of cyclic...

Great to have our manuscript with @sokrypton.org 's lab describing AfCyDesign finally out in @natcomms.bsky.social . Structure prediction, sequence redesign, de novo hallucination of cyclic peptides, and some binder design examples in this version.
rdcu.be/em0vA

1/ In two back-to-back papers, we present our de novo TRACeR platform for targeting MHC-I and MHC-II antigens

TRACeR for MHC-I: go.nature.com/4gcLzn5
TRACeR for MHC-II: go.nature.com/4gj5OQk

Hans Ellegren Welcoming David Baker at Stockholm Arlanda.

Today, several #NobelPrize Laureates arrive in Stockholm, warmly welcomed by Hans Ellegren. Here we see David Baker stepping off the plane at Arlanda.

This week is packed with inspiration, press conferences and lectures, so stay tuned! 🌟
@uofwa.bsky.social @hhmi.bsky.social
#Science #AcademicSky

Thanks for the shoutout! I agree - We have not tried it, but
don't expect it to work well for disordered targets with low-throughput testing. Not yet, at least! 😀

Hats off to the people compiling all those starter packs—it has made the move to this site so much easier!

Ha ha Probably not! :)

I may have figured out how to add a GIF of diffusion trajectory. Lets see! 😀
i.giphy.com/media/v1.Y2l...

It is a really fun time to be designing peptides/proteins. Please reach out if you have targets you would like to design macrocycle binders.

None of this would have been possible without all the great collaborators at @uwproteindesign.bsky.social and beyond (still trying to find everyone here!). There is much more to come as we continue to fine-tune and expand RFpeptides.

Perhaps the most fun part for us was the RbtA, where we did not have the target structure available when we designed against it. So we predicted the structure using AF2/RF2 and then designed against the predicted structures. Tested < 15 designs and got a Kd <10 nM binder!

X-ray structures for the macrocycle bound complexes also match very closely with the design models (CA RMSD < 1.5 angstroms). The designs are diverse: helix-containing (MCL-1/Mdm2), beta-strands (GABARAP), and loopy (RbtA).

We used RFpeptides to design binders against four different targets: Mdm2, MCL-1, GABARAP, and RbtA. For each of the targets, we experimemtally tested <20 designs. We got 1-10 micromolar Kd binders against Mdm2 and MCL-1, and 1-10 nM binders against GABARAP and RbtA.

Here, we modified RFdiffusion positional encodings to design cyclic peptide backbones against selected targets, followed by sequence design using ProteinMPNN. Final designs were selected based of confidence metrics from AF2/RF2 re-prediction and Rosetta-based interface quality metrics.

So the design pipeline has to be good at designing and also selecting the best 10-20 binders. And it should also work for diverse targets. RFpeptides seems to be able to address a lot of those early issues and meet the requirements.

Why are we excited about it? Well, we spent a lot of effort over the years to accurately design high-affinity binders with our physics-based methods without much success. Since we rely on chemical synthesis of macrocycles, we were limited to making and testing only 10-20 designs/target in our lab.

Here goes the skeetorial for the latest preprint from our lab describing RFpeptides, a pipeline for design of target-binding macrocycles using diffusion models. Big shoutout to Stephen Rettie, David Juergens, Victor Adebomi for leading the project (1/n)
Preprint link: www.biorxiv.org/content/10.1...

Here is a GIF in the meantime:

Accurate de novo design of high-affinity protein binding macrocycles using deep learning The development of macrocyclic binders to therapeutic proteins has typically relied on large-scale screening methods that are resource-intensive and provide little control over binding mode. Despite c...

Super excited to share the latest preprint from our lab on macrocycle binder design! Skeetorial (or whatever they are called) to follow soon. Thanks to all the amazing collaborators!
www.biorxiv.org/content/10.1...