Erzsébet Fichó retweeted
a new AlphaFold model: in AI for biology, you get what you train for
since the release of AlphaFold in 2020, a once-hypothesized revolution in drug discovery has failed to materialize. while accurate at predicting protein structure, the utility of AF for drug discovery has been limited. new AF models published this week by isomorphic labs change this.
a useful case study for understanding the difficulty of translating protein structure algorithms like AF to drug development is molecular docking. docking is a computational method for predicting the orientation and position of a ligand when it binds to a protein with the goal of generating new small molecule drugs, and is the workhorse of structure-based drug design. the docking process requires the crystal structure of a protein bound to ligands, however these structures are expensive/difficult to generate.
there was a lot of excitement after AF’s release about the replacing crystal structures with AF in docking in order to discover drugs more quickly and cheaply. many expected this to work because AF-generated structures closely match the crystal structures of protein binding pockets. however, several recent papers [1-3] have found that despite AF’s ability to predict protein structure, the accuracy of molecular docking decreases significantly when using AF-generated structures.
why didn't AF-generated structures translate to an ability to predict drug binding? generating a structural model for a small molecule binding to a protein turns out to be a much more difficult problem than protein structure alone. the space of possible small molecules (10^60) and structures is much larger than that of proteins. further, ligand binding changes protein structure, and AF wasn't trained on ligand-protein complexes.
a white paper out this week [4] describes a new iteration of AF that achieves stronger performance on molecular docking. the paper is intentionally short on methodological details, but from what we can tell, new AF models were trained on protein complexes with non-protein elements, including small molecule ligands. the model achieves SOTA accuracy on molecular docking benchmarks, outperforming both non-ML and ML based methods. importantly, comparison methods used ground truth bound protein crystal structures as input.
what’s the takeaway? one important, obvious lesson is that in AI for biology, you get what you train for. to predict ligand-protein interactions, you must train on train on ligand-protein complexes. to predict antibody-antigen interactions, you must train on antibody-antigen complexes. simply training on proteins (first generation of AF) is not sufficient for many applications in drug discovery, where the goal is to model complex interactions between many biological molecules. the capabilities of AI models in biology will only continue to advance as we incorporate specific and relevant biological data into model training to bridge the gap towards real-world impact.
@maxjaderberg @tfgg2 @demishassabis @IsomorphicLabs
Erzsébet Fichó retweeted
BREAKING NEWS
The 2023 #NobelPrize in Physiology or Medicine has been awarded to Katalin Karikó and Drew Weissman for their discoveries concerning nucleoside base modifications that enabled the development of effective mRNA vaccines against COVID-19.
Erzsébet Fichó retweeted
Meet @theannalytical, the @Google developer and drag queen who makes coding glamorous, accessible, and, of course, fun.
Anna Lytical also aims to improve LGBTQ+ representation by uplifting the marginalized voices in tech.
yahoo.com/now/anna-lytical-d…
Erzsébet Fichó retweeted
please do not share unverified footage/info from ukraine. russia maintains intense disinfo attack to fuel panic and fear. ukraine govt is doing superb job informing the public.
WAIT FOR VERIFIED INFO PLS
Erzsébet Fichó retweeted
📢 | We have received this letter from Town fan Joe, aged 6 and a half.
We'd really love to get in touch with Joe, but we don't have a return address.
If anyone recognises the writing or thinks they know who Joe is, please email [email protected]
#STFC 🔴⚪️
Erzsébet Fichó retweeted
A foxpost beállt a #nemluxustáska kampányba: aki #nemluxustáskát szeretne küldeni, elég összeraknia az adományt egy táskába és ingyen berakhatja egy foxpost automatába, és ők elviszik a gyűjtőhelyre ❤️❤️❤️ zseniális, szuperjó, nagyon menő ❤️❤️❤️#foxpost #nemluxustáska
Erzsébet Fichó retweeted
Some people search all their lives for it, but the meaning of life is literally just in the dictionary? 🤔
Erzsébet Fichó retweeted
My greatest inspiration. It was the way he raced, his passion for life and for the sport. But more than anything it was the way he faced alone a system that wasn’t always kind to him. May Aryton Senna’s legacy live on forever, especially vibrant this weekend
Erzsébet Fichó retweeted
This is the moment at Budapest Airport when the Pope makes his favourite 🇭🇺 Hungarian joke with me:
"Why will we speak Hungarian in heaven? Because it takes an eternity to learn it!"
Hence me laughing my head off.
Erzsébet Fichó retweeted
the outpouring love & support I’ve received has made me realize I’m more than my accomplishments and gymnastics which I never truly believed before. 🤍
Erzsébet Fichó retweeted
If this gets 20k likes, we’ll replace the paperclip emoji in Microsoft 365 with Clippy.
Erzsébet Fichó retweeted
Happy birthday, @captaintommoore!🎉
The WWII veteran has received over 120,000 cards for his 100th birthday after raising over £29 million for Britain’s NHS. chdr.tv/1028a
Erzsébet Fichó retweeted
“Hidden Figures” reminds us that Computers & Calculators were smart women who knew math. In the early US Space Program, that math was orbital mechanics, led by Katherine Johnson 1918-2020. May the number of those inspired by her story be computationally incalculable. Godspeed.
Erzsébet Fichó retweeted
Well done @Manu___Singh and Joanne! It was a pleasure to contribute to this amazing study!
Lymphoma Driver Mutations in the Pathogenic Evolution of an Iconic Human Autoantibody cell.com/cell/fulltext/S0092…




































