Felix Pultar retweeted
Frontier LLMs are getting better at chemical synthesis planning, but RetroChimera - our specialised model - is still top-1 😎
(and unlike GPT-5.6 Sol, you can run it locally in <16GB of GPU memory)
Big update to the Insilico #DDDBenchmark (Drug Discovery and Development) portal.🧪
The portal now includes purpose-built specialist models alongside the latest frontier models, including Opus 5 and the GPT 5.6 family (☀️Sol, 🌎Terra, 🌓Luna), giving a more complete view of how general-purpose and domain-specific AI compare across drug discovery tasks.
The results are especially striking on our multi-step synthesis planning #URSAbench. GPT 5.6 Sol (High) is now the top-ranked LLM and ranks 🥉 #3 overall, behind only two dedicated retrosynthesis planners, @maggie_hott @OpenAI.
General-purpose models are closing the gap fast. But the specialist leader still sets the pace: #RetroChimera by @marwinsegler @MaziarzKris from @Microsoft remains the only model to break 30 points and continues to hold the 🥇#1 spot 💪. #AiZynthFinder by @AstraZeneca takes 🥈.
Finding a promising molecule is often not the hard part in chemistry. Figuring out how to make it is.
That's what our model RetroChimera goes after. Published in @Nature today, model and weights on GitHub (MIT).
nature.com/articles/s41586-0…
#AIforScience #DrugDiscovery
1
6
149
Thank you for the great work together @marwinsegler @MaziarzKris @fiberleif @tobiasgensch @JunrenLi80950 @austinjtripp @jla_gardner @hubert_misztela @rishirg and our collaborators at GSK and Novartis
4
114
Felix Pultar retweeted
Synthesis is a major bottleneck in drug discovery. AI synthesis planning can help propose better routes and make generative molecular design tractable.
With our retroChimera model we have been pushing the boundaries of these models further - out in @Nature today. 1/2
Published this week, our review articles on machine learning interatomic potentials in @CHIMIA_journal with @rinikerlab : doi.org/10.2533/chimia.2026.… we put special emphasis on ML/MM approaches that further reduce computational costs compared to DFT #compchem #ml
2
84
Ever wanted to run MD simulations of entire proteins in water with DFT accuracy?
Meet AMPv3-BMS25, the latest iteration of our AMP multiscale neural network potential by @rinikerlab.
Read more in the preprints: doi.org/10.26434/chemrxiv-20… doi.org/10.26434/chemrxiv-20…
1
93
We showcase its scalability and accuracy across various benchmarks (>23 μs!), modeling solvation free energies, protein structural features, and free-energy profiles of enzymatically catalyzed reactions.
1
36
We have made the code, weights, and training dataset freely available for the community.
#ComputationalChemistry #MolecularDynamics #MachineLearning #StructuralBiology #DFT #AI #NNP #MLIP
31
Happy to see our latest interdisciplinary work on mutanobactin D published in @acsJACS. I am truly grateful to have worked with so many talented people! Check out the paper to see our combined experimental and computational work: pubs.acs.org/doi/10.1021/jac…
60
Happy to see our graph neural network implicit solvent model out in JCTC today. Check it out if you want to model explicit solvation effects in your QM calculations but want to keep computational costs low: pubs.acs.org/doi/10.1021/acs…
50
Excited to see our new paper published:
pubs.acs.org/doi/10.1021/jac…
Great collaboration with Moritz Thürlemann, who helped me get started with neural network potentials. Thanks to Sereina Riniker @rinikerlab for giving me the freedom and guidance to explore this exciting project.
47
Felix Pultar retweeted
Our latest work on new azoacetylene chemistry is now out in @J_A_C_S: straightforward and mild access to complex azotriazole photoswitches with cool properties, also implemented for generation of photoswitchable conjugates. Great teamwork! #OpenScience pubs.acs.org/doi/10.1021/jac…
Thank you @DirkTrauner and Andrej Shemet for highlighting our
@J_A_C_S paper on mutanobactin in #synfacts @thiemechemistry
thieme-connect.com/products/…
@CarreiraGroup @rinikerlab @LabLeibundgut
Felix Pultar retweeted
Our adventures with THC continue... Thanks to @SchafrothMA and @gertschgroup and many others, we were able to identify and evaluate the cannabinomimetic effects of Δ9-cis-THC stereoisomers from natural cannabis sources.
pubs.acs.org/doi/full/10.102… @ACSPublications
Felix Pultar retweeted
Serena Riniker and her research group @rinikerlab @ETH_DCHAB are using computer simulation to support the rational design of permeable cyclic peptides doi.org/10.2533/chimia.2021.…
Felix Pultar retweeted
Flow chemistry is also contributing greatly to peptide and protein synthesis as researched by @NinaHartrampf and Kevin Schiefelbein @UZH_Chemistry
doi.org/10.2533/chimia.2021.…
Felix Pultar retweeted
...organic and physical chemists working together on a natural product (!): configurational assignment by IR spectroscopy thanks to collaboration with Riniker Group @rinikerlab pubs.acs.org/doi/10.1021/jac… (2/2)
Felix Pultar retweeted
Mutanobactin D from the Human Microbiome: Total Synthesis, Configurational Assignment, and Biological Evaluation pubs.acs.org/doi/10.1021/jac… Our work on a natural product of the human oral microbiota: synthesis-biology-computation @rinikerlab @LabLeibundgut @FelixPultar ... (1/2)
My last day in the lab and our paper on mutanobactin D was published today in @J_A_C_S! Check out our work at pubs.acs.org/doi/10.1021/jac…! Thank you, @CarreiraGroup, @rinikerlab, @LabLeibundgut, and Schäffer lab for making this interdisciplinary work possible!







