Save the date: April 9 from 4pm to 6pm CET. Our department is hosting an online seminar with @noeliaferruz.bsky.social @sdomcke.bsky.social @const-ae.bsky.social who will talk about models for protein design, large-scale perturbation screens, and benchmarking of perturbation prediction models.
The CRG PhD call is now open. Exciting opportunities across diverse topics and within a world-class scientific environment.
Our group is offering one PhD position to study chromatin evolution.
Consider applying or share with anyone who might be interested!
www.crg.eu/en/content/t...
Protein Structure Evolution (ProSe) Seminar is now on BlueSky! Every 2nd Tuesday, 4PM GMT.
Sign-Up: tinyurl.com/prose-seminar2
Organized by Claudia Alvarez Carreño, Zachary Ardern, Lars Eicholt, Carolina Sanchez-Rocha,
Sergio Romero Romero and Md. Hassan Uz-Zaman.
Are you looking for a PhD? Join us in Barcelona! You'll dive into a community of >100 PhD students from 30 countries exploring the frontiers of biology. You can also join an online workshop on 6 November (15:00 CET) to learn how to find the right lab for you.
More info: www.crg.eu/en/content/t...
Dimitrije Ivančić (co-first author of the paper), Marc Güell (ICREA researcher and CSO of Integra Therapeutics), Avencia Sanchez-Mejías (CEO of Integra Therapeutics) and Alejandro Agudelo (co-first author of the paper)
#Research #Biomedicine
Generative AI is more efficient than nature at designing proteins to edit the genome 🧬💻✂️
This is the conclusion of research published at @natbiotech.nature.com by @marcguellc.bsky.social Integra Therapeutics & @noeliaferruz.bsky.social - @crg.eu
📰 tuit.cat/n7ap6
Thank you very much ☺️
This work has been possible thanks to an incredible team effort w/ Alejandro, @marcguellc.bsky.social @noeliaferruz.bsky.social and the integra therapeutics team!
my father-in-law is a “I know a guy” Guy. sadly all the old Guys are aging out of the Guy Economy. Guys are fundamentally incompatible with Hustle Culture because it’s not about “winning” a deal, it’s about collecting favors and goodwill in a mutually-beneficial cycle. protect your local Guy Economy
👀
CATH turns 30 years old this year!
We are organising a 1-day symposium on September 16th at UCL, highlighting recent AI-based developments to enhance protein family classifications, annotations and analyses.
www.eventbrite.co.uk/e/protein-an...
Woke up this morning in Greenland. It is game time for my cameras. National Geographic Endurance safely crossed the Denmark Straight and into this fjord! #EastCoastKin #photography #Greenland
This preprint from Helen Sakharova is one of the coolest things to come out of my lab: “Protein language models reveal evolutionary constraints on synonymous codon choice.” Codon choice is a big puzzle in how information is encoded in genomes, and we have a new angle. www.biorxiv.org/content/10.1...
Figure 1
Figure 2
Figure 3
Figure 4
Evaluating zero-shot prediction of protein design success by AlphaFold, ESMFold, and ProteinMPNN [new]
Zero-shot protein design assessment filters, but accuracy predicting success is limited.
But independent of this, what I find is that since Bill Degrado coined the term in the 80s, the field has evolved so much that it no longer captures today’s methods.
E.g in this (all-time favourite!) review, Baker and Po-ssu define de novo in a way that would exclude most AI methods today:
That’s an interesting paper, I must read 🙂
A bit off-tangent now, but I forgot to mention that Pubmed isn’t normalized, and all terms look as they are increasing exponientially, see for example banana 😄 (1/n)
🚀 We’re excited to share our latest publication in collaboration with Verena Ruprecht's group at @crg.eu
Mechanobiology meets Nuclear Metabolism
www.nature.com/articles/s41...
Interestingly, diffusion models can naturally sample proteins with 0% seq identity!
In any case, last time I said we had generated de novo sequences with a pLM a bunch of reviewers avalanched toward my paper with knives and axes and I had to use “artificial”🤣
Before AI we reserved that term for proteins with no evolutionary history, i.e., designed from first principles, and consequently often had sequence identities ca. 0%. pLMs, unless steered, tend to naturally sample at >30-40% seqid and rarely explore novel folds.
Thank you everyone for the congratulations and the large number of applications! I want to specifically encourage applications for these positions:
* Postdoc in structural biology (cryoEM, ideally with X-ray too)
* Full time lab manager/technician with wetlab experience
Uncancelled! 🥳
🚨🚨🚨
Please RT!
We're looking for a postdoc to join an exciting joint project between our lab @upf.edu & @crg.eu (Barcelona) and the Sander lab @mdc-berlin.bsky.social (Berlin) investigating how alternative splicing and microexons influences the maturation of pancreatic islets.
Deadline: 30/09/25👇
Join Martin’s new lab, great location, super cool project, and the best advisor! What’s not to like?
Wohoooo 😍😍😍🚀🚀🚀
Struct bio ms titles over the years
1965 - A three-dimensional model of X
1985 - The structure of X
1995 - The structure of X reveals Y
2005 - Structural basis of
2015 - The molecular mechanism of
2025 - Harnessing AlphaFold
The internet remains undefeated 😂😅
We are looking to recruit a tenure track group leader in the field of Chemical/Synthetic Biology (in the broadest sense) to lead a research program within the Division of Protein & Nucleic Acid Chemistry ( www.jobs.ac.uk/job/DNW809/r... ) at the MRC Laboratory of Molecular Biology (LMB).
What
Congrats Angel!!😍
Aaah I could speak about this for hours😄. For PLMs trained with autoregressive task, I’d say they suffer from the two, with the latter being most problematic because they fail to generalize even in regions that are close by; except when we steer/reinforce them (requiring considerable manual tuning)
