Excited to share some recent work with @manntis4 on using ML to predict the aqueous solubility and pH-dependent solubility of unseen organic molecules.
This is an important and difficult task, as @wpwalters (and many others) have pointed out:
practicalcheminformatics.blogspot.com/2018/09/predic…
New Rowan releases today!
(1) We're very excited to be launching new aqueous-solubility-prediction models today, and to be adding pH-dependent solubility to the macropKa workflow (for subscribers).
Automatically finding transition states from reactants + products can be powerful, but it's easy to go wrong: many algorithms are unreliable or far slower than a scan.
We've worked hard to make Rowan's TS-finding methods fast & robust, and we're very happy with the result:
Automatically finding transition states from reactants + products can be powerful, but it's easy to go wrong: many algorithms are unreliable or far slower than a scan.
We've worked hard to make Rowan's TS-finding methods fast & robust, and we're very happy with the result:
Today on Rowan, we're launching the ability to run double-ended TS searches using the freezing string method (FSM)!
This makes it significantly faster and easier to find TSs—just draw the input and output structures, select the level of theory, and Rowan will do the rest.
200% faster inference for molecular simulation — meet OrbMol.
OrbMol is built specifically for molecules, from metal complexes to biomolecules and electrolytes.
It delivers state-of-the-art accuracy, including protein-ligand binding energies, experimental accuracy in molecular…
“We hold these truths to be self evident…” followed by a list of some of the most radical claims ever made, some of which seem dubious even to this day was an incredible move
We thought these TSs looked cool, so we recomputed one using UMA-S. With GPU acceleration, the entire TS (re)optimization and frequency calculation took.... 3 minutes.
(There's a lingering 2i cm-1 freq which could doubtless be removed for publication.)
labs.rowansci.com/public/calcula…
If you've been running Boltz-2 jobs yourself or through Rowan, you may have noticed that lots are failing this week!
We've been tracking issues with the default Boltz-2 MSA server and working to fix this for our users. Here's our full incident report:
rowansci.com/blog/msa-failu…
If you are running large batches of co-folding jobs and are affected by changes to the ColabFold MSA server at api.colabfold.com, we'd love to be a part of your solution.
If you are running large batches of co-folding jobs and are affected by changes to the ColabFold MSA server at api.colabfold.com, we'd love to be a part of your solution. https://t.co/W08K1qhAhh
Our intern @ishaanganti joined us this summer to work on one of the biggest areas in computational chemistry: protein–ligand binding-affinity prediction.
There's a big gap between docking (fast) and FEP (slow); Ishaan worked in this gap on Pareto improvements to MM/GBSA & PBSA.
Heres a cool approach for determining uncertainty in NNP predictions by predicting with multiple output heads and taking the std
arxiv.org/pdf/2508.09907
Heres a cool approach for determining uncertainty in NNP predictions by predicting with multiple output heads and taking the std
arxiv.org/pdf/2508.09907
NNPs are becoming very accurate for gas-phase small molecules. But can current approaches scale to the condensed phase, or will scale inevitably create new challenges?
Today on the blog, we review how scientists are incorporating long-range forces into NNPs.
If you haven't checked the bottom of your Rowan logfiles, you're missing out!
Every successful workflow ends with a quotation, in keeping with the field's traditions (h/t @gaussianquotes). We add new quotes frequently.
Here are a few representative examples:
We bid farewell to two of our interns today. Here's what they had to say about their time at Rowan and what they learned (including advice for the daily company GeoGuessr game):
rowansci.com/blog/intern-re…
2K Followers 1K FollowingAI/ML, biotech, policy, flying, sometimes in that order | SovAI Unit @SciTechgovuk | PhD @Cambridge_Uni | Prev @benevolent_ai @imperialcollege | Glider Pilot
196 Followers 3K FollowingInvest lo que me rodea,
Salud Int., Permacultura, Gob-distribuida,Movimient opensource, Hackers sociales y Simpatizante ThevenusProject https://t.co/mtNp1Roxn7
100 Followers 2K FollowingAI/ML Engineer | AI x Healthcare | LLMs, NLP, Agents | Building HealthTech AI | Open to collab & remote roles
LinkedIn ⬇
https://t.co/QPCmZMBPQ0
35 Followers 462 FollowingI'm solidifying my understanding of web dev fundamentals, and improving my React skills. I'm also learning game dev with Unity!
355K Followers 1K FollowingML/AI researcher & former stats professor turned LLM research engineer. Author of "Build a Large Language Model From Scratch" (https://t.co/O8LAAMRzzW).
1.4M Followers 1K FollowingBuilding @EurekaLabsAI. Previously Director of AI @ Tesla, founding team @ OpenAI, CS231n/PhD @ Stanford. I like to train large deep neural nets.
45K Followers 794 FollowingHighland Traveller descendant, 12+ years nomadic. Catholic.
I write for a living at Hickman's Hinterlands at https://t.co/czRSBad12f
23K Followers 2K Following@Penn mathematician; engineer; educator;
assoc. dean of undergraduate education @PennEngineers;
illustrator; animator; e/acta non verba
1.5M Followers 1K FollowingOpen Source Intelligence Monitor focused on Europe and Conflicts across the World. RT ≠ Endorsement. Want to Support my Work? https://t.co/PcUbewvWPr
365K Followers 6K FollowingChief Scientist, Google DeepMind & Google Research. Gemini Lead. Opinions stated here are my own, not those of Google. TensorFlow, MapReduce, Bigtable, ...
12K Followers 644 FollowingCEO @ Nomic. We build models that enable AI systems to operate in the hardest unstructured data domains.
https://t.co/NbsLHLWfy8
5K Followers 7 FollowingFeed that aggregates publications about Machine Learning in Chemistry from over 200 journals of 15 editors. Contact @CYL_Lab for any question.