A huge thanks to the team: TJ Lane, Virginia Apostolopoulou and Jasper van Thor, and to @rs-station.bsky.social for supporting the project! You can check out the full paper here: www.nature.com/articles/s42... or try it on your data: github.com/rs-station/m... 8/8

We hope this tool becomes a useful addition for the community working on dynamic systems, transient complexes, or small fragment screening. 7/8

Why? Many experiments (e.g. time-resolved crystallography 🎥+ ligand screening 💊) generate minor populations that are hard to find and model. Our method gives a way to detect them more reliably. 6/8

Xtallography can’t measure phases, though they carry key info. With TV denoising as a Bayesian prior, we infer these latent phases. Inspired by coherent diffractive imaging, we embed the TV denoiser in an iterative EM loop that fixes experimental amplitudes, updating phases. 5/8

Rather than tweak parameters, we aimed for automation: optimal, reproducible, bias-free. Assuming noisy maps resemble Gaussian white noise, we use difference map negentropy–a measure of how non-Gaussian a signal is–as an automatic objective to optimize our parameters, avoiding manual tuning. 4/8

We apply TV denoising 📺 to difference maps, helping suppress noise and reveal subtle density changes corresponding to low-occupancy species. 3/8

🕺 As a retro detour from ML-based methods, we introduce total variation (TV) denoising to boost SNR in crystallographic difference maps. TV “flattens” regions like solvent or areas far from a ligand site while making minimal changes to the underlying data. 2/8

Write to us on GitHub for any issues and we'll improve! We are working to have a server available in collaboration with Phenix very soon.

@minhuanli.bsky.social and I are really excited to see new users trying ROCKET out. If you haven't heard about it but are curious, our GitBooks are a good place to start: rocket-9.gitbook.io/rocket-docs

At the moment we both cluster and uniformly sample and the validation of whether we have a better model compared from a full MSA comes from fit to the data, at the end, and geometric validation.

(1) ROCKET uses experimental information to score different conformations after subsampling (2) that first conformation arising from clustering is further modified through gradient descent in continuous MSA cluster profile space.

Thanks a lot Jake! The reliance of ROCKET on the MSA subsampling is different from other work in, I would say, 2 main ways.

Hey Martin, yes — www.biorxiv.org/content/10.1... it’s a little (too far) down the thread!

It has been very much on my mind – was excited to see the code posted!

Very grateful for the work, support, and guidance of all authors: Airlie, Tom, @randyjread.bsky.social, @hekstralab.bsky.social, and @moalquraishi.bsky.social. It’s a privilege to work with such a great team. 14/14

Very interesting work is also happening using diffusion-based priors! 🔗Solving Inverse Problems in Protein Space Using Diffusion-Based Priors arxiv.org/abs/2406.04239 & Inverse problems with experiment-guided AlphaFold arxiv.org/abs/2502.09372 13/14

All method details and equations in our preprint: 🔗 www.biorxiv.org/content/10.1... Code available soon! 12/14

ROCKET performs a new type of structure refinement by optimizing latent representations in evolutionary space. This unlocks possibilities for high-throughput ligand screening, assemblies solved at low resolution, and conformational landscapes – automation 🔜 new frontiers. 11/14