Matthieu da Costa

Mercredi prochain je serai de retour à Villejuif pour les 10 ans d’iGEM IONIS.

Table ronde sur la biologie de synthèse face aux enjeux climatiques que je partagerai avec Andrew Tolonen… mon prof en 2015 et advisor IGEM en 2014.

Excited to announce a preprint describing our software package Meeko! Meeko is a Python package that uses RDKit for receptor and ligand preparation, including protonation, bond order, and connectivity and processing of docking results. It is customizable and suitable for high-throughput workflows.

New Practical Cheminformatics Post
patwalters.github.io/Three-Papers...

With the release of Boltz 2, is there already a review comparing the efficiency of different (new) docking tools? (I’ve already tried Pocket Vina, which is actually quite good for high-throughput)

Entre deux simulations, épris de rationalité et mu par le désir de progrès, je me plonge dans cet ouvrage qui saura sans doute tempérer mes ardeurs…

The open-source Masala software suite: Facilitating rapid methods development for synthetic heteropolymer design Although canonical protein design has benefited from machine learning methods trained on databases of protein sequences and structures, synthetic heteropolymer design still relies heavily on physics-b...

We have a new preprint out on the Masala software suite, a free and open-source platform for easy biomolecular modelling methods development. www.biorxiv.org/content/10.1...

I'm excited to share our new preprint: PocketVina — a fast, scalable, and accurate multi-pocket molecular docking method.

Docking remains essential in early-stage drug discovery, but recent deep learning–based approaches still face limitations in generating...

Thread - (1/n)

Screenshot of PrankWeb.

🚀 Our paper on the updated PrankWeb, a web app for protein binding site prediction using P2Rank, is out in
@narjournal.bsky.social. Now with docking integration!

👨‍💻 Try it: prankweb.cz
📄 Paper: doi.org/10.1093/nar/...
💻 GitHub: github.com/cusbg/prankweb

Developed at @cusbg.bsky.social

The heptagonal box model is a specificity and activity model that helps identify the type of reaction an enzyme catalyzes and its substrate, but it is not meant to indicate how well the enzyme performs. Structural features clearly play a critical role in catalysis (dynamics, binding, etc.).

It actually feels like the opposite here in Paris. I keep seeing more and more people with a book in hand on public transport. Even at Abolis, there's a growing habit of sharing and recommending books.

EvoWeaver: large-scale prediction of gene functional associations from coevolutionary signals - Nature Communications Despite having structures for most proteins, we still do not know their function. Here, authors introduce EvoWeaver, a de novo method to identify genes working together using evolutionary information....

Proud to announce a new publication from the lab: EvoWeaver: large-scale prediction of gene functional associations from coevolutionary signals
www.nature.com/articles/s41...

Close to recreating it in ChimeraX:

1. Install the python file: github.com/smsaladi/chi...
2. rainbow palette magma
3. col modify #1 blackness - 40
4. col modify #1 whiteness - 40
5. col modify #1 saturation - 50
6. col modify #1 lightness + 20

Room for improvement, needs more desaturation

Introducing CAMEO Structures & Complexes - automated weekly blind benchmarking of structure prediction servers. Now with heteromeric and protein-ligand complexes.
Join us and register your server now!
cameo3d.org

"De novo prediction of protein structural dynamics"

I'll be presenting an overview of the field tomorrow at a workshop. Link to a PDF copy of the presentation: delalamo.xyz/assets/post_...

Boltz-1: Democratizing Biomolecular Interaction Modeling Understanding biomolecular interactions is fundamental to advancing fields like drug discovery and protein design. In this paper, we introduce Boltz-1, an open-source deep learning model incorporating...

Boltz-1 update Boltz-1x introduces inference-time steering for much higher physical quality, CUDA kernels for faster, more memory-efficient inference and training.
www.biorxiv.org/content/10.1...

AFESM: a metagenomic guide through the protein structure universe! We clustered 821M structures (AFDB&ESMatlas) into 5.12M groups; revealing biome-specific groups, only 1 new fold even after AlphaFold2 re-prediction & many novel domain combos. 🧵
🌐 afesm.foldseek.com
📄 www.biorxiv.org/content/10.1...

Gene synthesis is often the most expensive part of protein engineering with generative models.

Happy to have played a small part in this work, where Chase developed a method for precision library construction at scale, with per-gene costs as low as $1.50.

@philromero.bsky.social

Three BioML starter packs now!

Pack 1: go.bsky.app/2VWBcCd
Pack 2: go.bsky.app/Bw84Hmc
Pack 3: go.bsky.app/NAKYUok

DM if you want to be included (or nominate people who should be!)

RFDiffusion2: a diffusion model that scaffolds enzyme active sites at the atomic level enables the generation of functional de novo enzymes

www.biorxiv.org/content/10.1...

Small proteins can be more complex than they look!

We know proteins fluctuate between different conformations- but by how much? How does it vary from protein to protein? Can highly stable domains have low stability segments? @ajrferrari.bsky.social experimentally tested >5,000 domains to find out!

Emergence of fractal geometries in the evolution of a metabolic enzyme

For future reference, this is the slide (without the bottom cropped) that should have been posted here. Boltz-1 and AF3 are actually performing similarly on the Chymase and Mpro datasets, and differ only on the Autotaxin dataset for still unknown reasons.

Modelling and design of transcriptional enhancers - Nature Reviews Bioengineering Enhancers are genomic elements critical for regulating gene expression. In this Review, the authors discuss how sequence-to-function models can be used to unravel the rules underlying enhancer activit...

We wrote a review article on modelling and design of transcriptional enhancers using sequence-to-function models.

From conventional machine learning methods to CNNs and using models as oracles/generative AI for synthetic enhancer design!

@natrevbioeng.bsky.social

www.nature.com/articles/s44...

GitHub - baker-laboratory/PLACER: PLACER is graph neural network for local prediction of protein-ligand conformational ensembles. PLACER is graph neural network for local prediction of protein-ligand conformational ensembles. - baker-laboratory/PLACER

We are happy to share that the code and weights of PLACER (formerly known as ChemNet) are now public!
This method allows for rapid evaluation of protein side chain and ligand conformational ensembles, with uses in ligand docking and enzyme evaluation.
github.com/baker-labora...