A new version of the Hitchhiker’s Guide is up, including general improvements: - a blurb on frame-based invariant GNNs (like AF2’s IPA) - new citations for unconstrained GNNs - clarifications on the universality of GemNet - and this hand-drawn figure PDF: arxiv.org/abs/2312.07511

Our Nature Reviews Physics perspective article on neural operators and their ability to accelerate simulations and design is now out. rdcu.be/dD8BI nature 1. Neural operators learn mappings between functions, e.g. spatiotemporal processes and partial differential equations.

If you are (or you know) a young researcher interested in doing a post-doc in Chile, mixing probability+graphs+deep learning, please contact me! Please spread this for visibility! (adding some images taken in the last 3 months for motivation)

Information geometry and optimal transport are cool in theory, but flow matching in the Fisher-Rao metric is even cooler in practice ! It was great to be part of this awesome team, thanks for having me!

sad but obvious realization, in the context of the recent gzip ~ llm scaling laws paper. LZ compression algorithms are optimal in the same way memorization is learning—only true with infinite data.

Simon Barnett Honest question. Do we mean dynamics as in time-dependence, or as in conformational flexibility and variability ? Both are important, but in a sense, different things.

Amazing first week of #MoML2024 🇨🇦

A 2D version of “AlphaFold pseudo Multi Dimensional Scaling” skills

Yesterday, I had the chance to present my first spotlight poster at #MoML2024 at Mila - Institut québécois d'IA AI Institute! I presented joint work done with Luisa Kärmer and our amazing advisors, Julia Westermayr and Jens Meiler Lab. Huge thanks to @Esporascicomm for the amazing graphic design work!

Had a great time at the ML for Drug Discovery Summer School & MoML conference in Montréal (my first time in🇨🇦!) Kudos to the organizers for arranging such a great lineup of talks & labs! thanks to the hackathon, got some hands on experience with some concepts taught👨🏼‍💻

We recently created some new tools in Rosetta around ML protein design methods (including ProteinMPNN, ESM2, MIF-ST). You can run all of this using `docker run -it rosettacommons/rosetta:ml` (no python involved). We then benchmarked the different methods: biorxiv.org/content/10.110…

Amazing time at the I-X open day - super cool facilities and great content😆. Excited to receive a poster prize at the end! Electronic aspects inspired from Brady Johnston who gave great advice to make sure the iPad stayed stuck! Animation code taken from Sergey Ovchinnikov 😎

NEWS | Epic announcement coming tomorrow, thanks to Rosetta Commons Jens Meiler Lab 🎉 Stay tuned!

4/ Metagenomes feature significant bias and redundancies, but deduplicating large genomic databases can be computationally expensive. We implement *genomic SemDeDup*, an embedding-based deduplication for genomic sequences, resulting in tunable balancing and pruning of the corpus

7/ In particular, we showcase gLM2's ability to directly learn coevolutionary signal in protein-protein interfaces with no supervision! The learned contact maps can be extracted using Zhidian Zhang et al's categorial Jacobian method.

Kevin K. Yang 楊凱筌 x.com/HattoriLab/sta…

Started teaching again! This time decided to try use #claude (Anthropic) and CodeSandbox for hosting to implement an interactive GREMLIN (Potts) model (w = coevolution, b = conservation) to show students how you go from MSA to contacts! 9kssnq.csb.app

Thrilled to have shared my personal and scientific journey on the Baker Lab podcast! I hope it serves as an inspiration for aspiring scientists in Latin America.

Yuri Deigin OpenAI Retro Biosciences For every new technology (in this case engineered proteins) you have to first make a PoC.

Excited to share work with Zhidian Zhang, Milot Mirdita, Martin Steinegger, and Sergey Ovchinnikov biorxiv.org/content/10.110… TLDR: We introduce MSA Pairformer, a 111M parameter protein language model that challenges the scaling paradigm in self-supervised protein language modeling 🧵