Really interesting paper: ESM-all atom (ESM-AA) to compete with structure-focused Rosetta-AA and AlphaFold3! ✨ ESM-AA uses multi-scale code-switching to model molecules at residue and atom scales, using multi-scale position encoding to capture contexts in both regimes. It seems

My new novel, DARKOME, a near-future biotech thriller, will be published in the UK on September 5th 2024. Pre-order it here: gollancz.co.uk/titles/hannu-r… For a taste, here’s the prologue, read by me: youtu.be/rR77_JTtq5Q?si… A thread on how the book came to be!👇🏻

OpenFold paper is now out in Nature Methods. Timely. I’ll say more about AF3 at the end of this thread, but first I wanted to highlight what’s new since our initial release. (1/9)

Aviv Regev on collaborations: be generous with your ideas, your abilities, with resources that you have, your time, your mindset... And when everyone in a collaboration behaves in this way, then of course, it's a lot of fun.
podcasts.apple.com/us/podcast/avi… The Night Science Podcast

I’m speaking today at PEGS Summit, 5:10 pm, Eng Antibodies. Talk will cover new “solMPMP” tech from Nabla Bio where we use generative ML to create soluble proxies of valuable multipass membrane protein drug targets, and use these proxies to facilitate design of new biologics.

When you want to summon Michaelis and Menten

💯. Doing something interesting is inherently risky. Academia is actually largely designed to enable risks from a position of relative safety. The recent trend towards wanting everything laid out and planned is stultifying.

Strong disagree

Pssst want to work on aging 👀?
Weekly Retro is hiring tweet: We have open positions all over the company across science, operations (front office, lab management, facilities), including one ML Engineering position

I am delighted to share the newest work from the Sternberg Lab, in which we discover RNA-templated gene creation as a mechanism of antiviral immunity. This is a story for anyone who loves molecular biology, and one that repeatedly left us in disbelief. doi.org/10.1101/2024.0…

Looks useful...

I agree with this prediction, but I think it gets causality backwards.

The curve of moores law was carefully choreographed between visionary buyers and fiercely competitive providers — the big use cases of today (streaming video, AI) were easily articulated in the 1960’s, and

Such a clever new LM architecture: Orchid uses data-dependent convolutions to overcome the quadratic complexity of traditional attention. 🌺 Achieves quasilinear scalability with dynamic kernel adjustments for long sequences! Also, super strong as a BERT model!

Paper:

Humans: nanotechnology will never work!!!

Bacteria: hold my beer.

Excited to share Taraneh's preprint: A complete map of specificity encoding for a partially fuzzy protein interaction biorxiv.org/content/10.110…

Future of biotech discovery - in vivo multiplexed screening of 1000s of gene targets and therapies per animal - so excited for the progress Gordian Biotechnology has made so far...

chat, is this… real?
magnificent

Two exciting studies out in nature show that high tumor PGE2 limits IL-2 sensing and antitumor function in TIL. Disrupting this axis can potently enhance antitumor responses. Nice work George Coukos Jan P. Böttcher and teams!

nature.com/articles/s4158…

nature.com/articles/s4158…

How do we interpret molecular mechanisms of protein allostery from deep mut scan experiments? We have built a generalized thermodynamic model to infer and predict how mutations impact inter-and-intradomain allostery. Great collab with Qiang Cui BU
eLife - the journal doi.org/10.7554/eLife.…

ACGT Research Fellows, and members of the ACGT Scientific Advisory Council, Carl June, MD (Penn), and Michel Sadelain, MD, PhD (@MSKCancerCenter), received the Breakthrough Prize for Life Sciences on Saturday!

(1/3)