Huge thanks to the whole team: Montader Ali, Klara KropivΕ‘ek Brumat, Leonardo Salicari, Samo Miklavc, Aibek Kappassov, Chengcheng Fu, Pietro Sormanni, Ario de Marco, and Victor Greiff!
We also provide actionable guidance on sampling strategies, compute-accuracy tradeoffs, and confidence calibration relevant to pharma-facing AI biologics workflows
Our findings include:
- Confidence scores don't separate cognate and non-cognate complexes
- Geometric plausibility doesn't imply specificity
- Cross-tool agreement is weak
- Sampling improves geometry, not specificity or confidence alignment
- Compute and energy costs vary widely
We benchmarked publicly available AF3, Boltz-2, and Chai-1 across 106 experimentally solved VHH-antigen complexes and over 11,000 shuffled non-cognate pairings. We chose the VHH nanobody format for its single-domain simplicity (easy to interpreat paratope-epitope interactions)
We examined how well AI-based antibody-antigen structure prediction pipelines perform at the practical discovery-critical task of differentiating true binders from plausible but incorrect ones. Short answer: not very well.
Read the paper here: www.biorxiv.org/content/10.6...
I'm doing a market research about eating habits and appetite control. Are you hungry? π
If you have 2 minutes, please fill in the Google form below (anonymous, unless you want to have a follow up chat): forms.gle/Mpc86LbhGSyH...
Spread the word!
Thank you!
In 2019, we published a clustering & nomenclature of beta turns in proteins & found 18 unique types. The code was in python2. I've just rewritten it in python3, because it might still be useful to people.
The paper & code:
journals.plos.org/ploscompbiol...
github.com/DunbrackLab/...
Clustering hundreds of millions of molecules in a single workstation is now a possibility!
Check the new code: github.com/mqcomplab/bb...
A compelling review of how ML/AI could help in the quest to find an enzyme for every reaction.
@jsunn-y.bsky.social @francescazfl.bsky.social Yueming Long @francesarnold.bsky.social
www.cell.com/cell-systems...
Preprint: De-novo design of proteins that inhibit bacterial defenses
Our approach allows silencing defense systems of choice. We show how this approach enables programming of βuntransformableβ bacteria, and how it can enhance phage therapy applications
Congrats Jeremy Garb!
tinyurl.com/Syttt
π§΅
Glad to share the final version of our story about the UBR4 complex, an E4 ligase protein quality control hub @science.org. Now with more cryo-EM structures and a deeper dive into substrate recognition, especially escaped mitochondrial proteins @clausenlab.bsky.social www.science.org/doi/10.1126/...
#CCPBioSim will be running their annual training week of hands-on workshops in basic biomolecular simulation techniques from the 13th-17th October, in Sheffield.
For more details, visit www.ccpbiosim.ac.uk/training2025
The registration link there will open shortly! #compchem
New degrader paper using the immunoproteasome. Congrats Cody Loy and Tim Harris!
pubs.acs.org/doi/full/10....
encapsulate your target proteins for cryoEM structural determination within highly hydrophilic, structurally homogeneous, and stable protein shells π€©
www.biorxiv.org/content/10.1...
We're excited to present our latest article in Nature Machine Intelligence - Boosting the predictive power of protein representations with a corpus of text annotations.
Link: www.nature.com/articles/s42...
[1/4]
CACHE 7 is launched with support from the @gatesfoundation.bsky.social and unpublished data from Damian Young at @bcmhouston.bsky.social, Tim Willson @thesgc.bsky.social and Neelagandan Kamaria InSTEM. Design selective PGK2 inhibitors. We'll test them experimentally.
bit.ly/4lnVYOs
3D render of a Zincore complex locking ZFP91 onto DNA, promoting gene activation. Made in Blender with Molecular Nodes.
π’ Ready for this? A new way to think about Gene Regulation!
π¨A new coregulator complex, Zincore, acts as a "MOLECULAR GRIP", stabilizing TFs at DNA binding sites across the genome!π§¬
πͺProud to be part of this work led by @daaninthelab.bsky.social
Read more: science.org/doi/10.1126/science.adv2861
π My first Cryo-EM structures are OUT on the PDB!
πSuper excited about this milestone as PhD, itβs been a journey, and Iβm grateful to finally have these as my first Cryo-EM structures! π
πThank you for sharing!β€οΈ
Read @daaninthelab.bsky.social et. al.: www.science.org/doi/10.1126/...
β¬οΈ 2nd video
π¨Itβs out @embojournal.org π¨
Cardiolipin inhibits Caspase-4/11!
Huge thanks to my brilliant supervisors Kate & Pablo, to @s-burgener.bsky.social & Mercedes and all the @inflammasomelab.bsky.social, it was a joy working and laughing with you, and to @brozlab.bsky.social & Si Ming lab.
We have a new collection of protein structure generative models which we call Protpardelle-1c. It builds on the original Protpardelle and is tailored for conditional generation: motif scaffolding and binder generation.
Switching next to protein protein interactions, we used arguably my favorite protein A34F GB1 to demonstrate that these glycopolymers stabilize protein protein interactions in a similar fashion to protein folding, via chemical interactions.
Interested in solvation free energies of small molecules and proteins?
The latest @livecomsjournal.bsky.social tutorial by Egger-Hoerschinger et al provides a guide to quantifying hydration thermodynamics using Grid Inhomogeneous Solvation Theory (GIST):
doi.org/10.33011/liv...
#compchem
Pushing down to the structures of ever-smaller proteins (and ever-smaller crystals!)
Happy to share the latest from the lab, led by Daniel Alvarez, in collaboration with @lizconibear.bsky.socialβ¬. In this AA-MD tour-de-force, we delve deep into the mechanism and energetics of lipid uptake by bridge-like lipid transfer proteins, and we learn a few interesting things along the way...
Thrilled to share my first corresponding author paper as a project leader in @hillerlab.bsky.social βͺ Amazing work from first author @annaleder.bsky.social We discovered multi-chaperone condensates in the ER that revolutionise the current vision of protein folding!
www.nature.com/articles/s41...
Table of contents figure
Our latest Best Practices article "Developing Monte Carlo Methodologies in Molecular Simulations" is out now and describes how to derive acceptance probabilities for a variety of Monte Carlo moves:
doi.org/10.33011/liv...
#compchem
A benchmark dataset of 614 experimentally characterized de novo designed monomers from 11 different design studies shows that:
- deep learning structural metrics only weakly predict success
- The score distribution is different for different types of structures
@grocklin.bsky.social
Physics-based design of efficient Kemp eliminases
@lynnkamerlin.bsky.social
www.nature.com/articles/s41...
All all-atom diffusion model of protein sequences.
www.biorxiv.org/content/10.1...
Very excited about our latest all-atom generative model proteina, check out the project page (research.nvidia.com/labs/genair/...) and stay tuned for the code release soon!
