Olexandr Isayev 🇺🇦🇺🇸

Our latest @ChemRxiv preprint, "Transferable #MachineLearning Interatomic Potential for Pd-Catalyzed Cross-Coupling Reactions" Collaboration with @NSF_CCAS @LosAlamosNatLab @gabepgomes @bobbypaton #compchem #NSFfunded chemrxiv.org/engage/chemrxi…

@andrewwhite01 OpenEye had this feature for years

YES!! Finally a way to convert SMILES to IUPAC names. Ridiculous that this software has been missing for so long.

This Special Issue showcases advances in the creation of new frameworks for building more reliable Markov state models, allowing integration with experimental observables, and allowing the modeling of more complex systems in the future. Learn more: go.acs.org/eqz

Fable is so brilliant It can't answer any of my questions - cites the ai safety and biology concern ang goes back to opus 4.8 😫 (if you know you know 😉)

🚨Act now! If not now, when? This is an emergency. #science #academiclife #savescience science.org/content/blog-p…

@davidlmobley Yay! Congrats 🎉

Our paper on a first pass at automatically detecting slow torsional rearrangements in MD is out at JCIM: dx.doi.org/10.1021/acs.jc… No access? See the chemRxiv version: chemrxiv.org/doi/full/10.26… #compchem

Ukrainian 🇺🇦 embroidery, known as vyshyvanka - organic chemistry edition. #compchem #chemistry

2/2 The result is sobering: tool access helps, but is not enough. Even the strongest agents reach only 50.6% accuracy within 25% relative error on clean inputs, and performance drops under realistic noise. ChemCost exposes failures in grounding, retrieval, pack selection, arithmetic, and tool use. arxiv.org/abs/2605.07251

1/2 Can #LLM agents actually price a chemical reaction? #compchem In our new paper, we introduce ChemCost: a benchmark for chemical procurement cost reasoning across 1,427 reactions, 2,261 chemicals, and 230,775 supplier quotes. The task: identify chemicals, retrieve valid packs, normalize quantities, and compute cost.

Now as few weeks has passed, I still have #GenesisMission PTSD. However I ‘d like to acknowledge all research administration, OSR and support personnel. You are true heroes of this effort. 👏👏👏

Looking forward to this exciting event! Celebrating 35 years of outstanding Ukrainian 🇺🇦 chemistry @EnamineLtd . enamine-conferences.org

@katyenko @manntis4 @RowanSci @muni_bio @onepot_ai Yes, let me check if I can run Amber on comparable hardware for comparison

@olexandr good point, we updated the post. This was run on Nvidia L40S x4 using Rowan’s fast preset: 0.25 ns/edge, 24 λ windows, AM1-BCC. At 12 credits/min and $0.04/credit that’s $0.48/min, so ~$21.50 for the 45min run. Would love to see the setup + results from an RBFE run in Amber on 4x L40S, though. If it’s cheaper, faster and within the same accuracy, happy to do a post on it!

Traditional FEP is slow and prohibitively expensive, but @RowanSci FEP changes this. In this article, we show how rapid FEP can be placed earlier in the loop, connect to automated analogue enumeration in @muni_bio, and bring molecules straight into synthesis with @onepot_ai. muni.bio/research/fast-…

💯

HAPPY STAR WARS DAY! 🎉 #MayThe4thBeWithYou

@caco3cz Hmm… good question. I don’t know. There is no well defined learned representation (aka embedding) in force field. So this framework needs to be adapted somehow…

@olexandr may it work for classical force fields as well?

New preprint alert!🚨We introduce PROBE: a framework for uncertainty quantification in #machinelearning interatomic potentials (#MLIP). Instead of the typical regression approach, PROBE frames #uncertainty as a simple binary question: is this prediction reliable or not? This simplification makes the problem tractable, and we show that a lightweight classifier trained on top of a frozen MLIP's internal representations is all you need. Additionally, PROBE's attention mechanism produces per-atom importance maps highlighting exactly which atoms drove an unreliable prediction, at no extra cost. #comchem arxiv.org/abs/2605.00640

This is students dorm of my alma mater Dnipro National University today 💔💔❤️‍🩹 this is why ruzzia must be defeated as soon as possible… Donate to support Ukraine #StandWithUkraine️

@grantrotskoff @charles_irl Congrats. Great work 💪

Protein design has been dominated by diffusions due to a "structure-first" perspective. What about intrinsically disordered proteins? We scale language-based design using the modern RL stack and our model IDiom. Paper: biorxiv.org/content/10.648… Try it: idiom-designer.vercel.app

Joint JCIM & JCTC @JCIM_JCTC editorial: Advancing Reproducibility and Open Data in Theoretical and Computational Chemistry. #compchem pubs.acs.org/doi/10.1021/ac…