Gianni De Fabritiis

Office view

Preprint release 😀 of "Speak to a Protein," an AI co-scientist that facilitates data gathering and analysis in an interactive collaborative session. It is quite amazing to use. Preprint: arxiv.org/abs/2510.17826

Acellera/AceFF-1.1 · Hugging Face We’re on a journey to advance and democratize artificial intelligence through open source and open science.

Check out AceFF 1.1

huggingface.co/Acellera/Ace...

Heading to Boston for the free energy workshop...
www.feworkshop.org

First long waited summer day

📂 Stay organized while exploring molecules in open.playmolecule.com! Group them by clicking the Group button in the menu bar. Toggle the visibility of all systems in a group by clicking on it. Watch the video to see how grouping simplifies your molecular exploration!

🐍 Dive into molecule analysis with our Python terminal in open.playmolecule.com! Interact with loaded molecules using #Moleculekit. Execute Python code within the terminal to obtain information of your systems, apply filters, perform calculations, and more. Watch the demo!

Gianni De Fabritiis on LinkedIn: #ai #molecularreasoning #quantumchemistry #drugdiscovery I am really happy to announce our collaboration between Acellera Therapeutics and Psivant Therapeutics to push the frontiers of AI-driven drug discovery!…

I am really happy to announce our collaboration between Acellera Therapeutics and Psivant Therapeutics to push the frontiers of AI-driven drug discovery!
www.linkedin.com/posts/gdefab...

Interested in reasoning agents?

t.co/Ax9RtOLDCG

@andrew.diffuse.one

Acellera Therapeutics Unveils AceForce 1.0: A Novel Family of Next-Generation Neural Network Potential Accelerating Drug Discovery Acellera Therapeutics, a pioneer in computational chemistry and AI-driven drug discovery, today announced the launch of AceForce 1.0, its groundbreaki

Aceforce 1.0 is out and downloadable from Huggingface. It covers all interesting elements and charged molecules. Tested for relative binding affinity calculations. Just version 1.0, more to come.

www.businesswire.com/news/home/20...

BricksRL at Neurips. I am not going this year, but don't miss our spotlight paper/poster on RL with Lego for robotics.

Super excited to preprint our work on developing a Biomolecular Emulator (BioEmu): Scalable emulation of protein equilibrium ensembles with generative deep learning from @msftresearch.bsky.social ch AI for Science.

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

ACEMD v4 is out with the integration of neural network potentials. More to come soon.

O1 Replication Journey -- Part 2: Surpassing O1-preview through Simple Distillation, Big Progress or Bitter Lesson? This paper presents a critical examination of current approaches to replicating OpenAI's O1 model capabilities, with particular focus on the widespread but often undisclosed use of knowledge distillat...

I liked part I. Let's read part II.

arxiv.org/abs/2411.16489

How did you make that? Manually?

Marco-o1: Towards Open Reasoning Models for Open-Ended Solutions Currently OpenAI o1 has sparked a surge of interest in the study of large reasoning models (LRM). Building on this momentum, Marco-o1 not only focuses on disciplines with standard answers, such as mat...

Alibaba has their own version on GPT-o1. This might be the best description of “o1-type”systems so far arxiv.org/abs/2411.14405

After Martini, Espresso and Grappa methods/models in Italian drink names. Amaro was long overdue.

In Spring I come to ACS, you can get me a beer and we are even.

You are welcome!

The next one will be called Greg.

AMARO’s transferability was tested on unseen, fast-folding proteins like Trp-Cage and α3D. It effectively recovered free energy landscapes for most

This is version 1. Model fully available.

github.com/compsciencel...

AMARO leverages TensorNet, an O(3)-equivariant neural network, and omits hydrogen atoms to reduce complexity.

Trained on the mdCATH dataset, AMARO excels in scalability and generalization. It predicts forces with minimal error even for larger domains (>150 residues).

AMARO: All Heavy-Atom Transferable Neural Network Potentials of Protein Thermodynamics
A novel machine-learning-based force field for protein thermodynamics. It uses an all-heavy-atom approach without hydrogens to simplify simulations while maintaining key dynamics.
pubs.acs.org/doi/10.1021/...

Love to join in. thanks.

I'm making a list of AI for Science researchers on bluesky — let me know if I missed you / if you'd like to join!

go.bsky.app/AcP9Lix

Hi, please add me too.

OpenAI’s data scraping wins big as Raw Story’s copyright lawsuit dismissed by NY court The dismissal of Raw Story’s lawsuit is more than a win for OpenAI—it’s an indicator of how courts may handle similar copyright claims.

Raw Story's lawsuit against OpenAI for training on their news stories was dismissed by NYC court. Court said Raw Story didn't have standing because they couldn't prove injury venturebeat.com/ai/openais-d...

I went wing foiling at the Keys, last weekend. This was taken from the hotel window just after I woke up.