Making molecules is hard. How can we simplify the predicted synthesis routes of generated, property‑optimized small molecules? Our last work presents a framework that lets you "mix-and-match" multiple reaction constraints in the synthesis of your generated molecules.

Out now! @pschwllr, @SchwallerGroup, @loic_roch, @VictorSabanza and colleagues provide guidelines and recommendations for when to use multi-fidelity Bayesian optimization over their single-fidelity counterparts https://t.co/vFsEPABa0Q 🔓 https://t.co/mdALeF6amU

Coauthors: @JeffGuo__ , @ZJoncev, @JLuterbacher, @pschwllr Thanks to @NCCR_Catalysis and @EPFL_ReO GlobaLeaders for the support!

Using a generalist molecular generative model and reinforcement learning, we can address synthesizability multi-parameter optimization objectives without additional inductive biases. Code: https://t.co/jY6bWL1hNw Preprint:

We show the applicability of the framework in different use cases, related to drug discovery, industrial byproduct valorization and ultra-large-scale virtual screening.

✅ Enforce the presence of specific reactions in the synthesis or only certain reactions ❌ Avoid specific reactions 🟢 Enforce the presence of certain building blocks in the synthesis ⬇️ Minimize synthesis route length

Finally out! We have been working on a generative molecular design framework to allow steerable and granular control over synthetic routes for property-optimized molecules. ⚗️ 🖥️ Take a look below! 👇

Hello from Singapore 🇸🇬! Thrilled to be at #ICLR2025 presenting our work on fragment-based drug discovery 🧩. We go beyond virtual screening with a generative, structure-aware approach. 📃 https://t.co/rH6SkSCPIY 🔗 https://t.co/H8BTPGLZp8 A thread 🧵👇

Check out the updated published version of our pre-print in @ChemicalScience! (1) *General-purpose* generative model + retrosynthesis model = design molecules with optimized properties with an *explicit* predicted synthesis pathway Paper: https://t.co/y0H2wovtlF (1/2)

🎉Our BoLudo paper is in @J_A_C_S! Bayesian Optimization for nanocrystaL strUcture Design Optimization (jk, it's BOjana & LUDO😂) We show you can optimize ANYTHING when put on the right scale - even nanocrystals with surface energy! Also we discovered a new Cu shape! See how👇

🧵 Excellent showing from @SchwallerGroup at #NeurIPS2024! Our team received multiple spotlight presentations and acceptances across key AI+Science workshops, showcasing cutting-edge work in AI for chemistry and materials science.

Atinary @ #NeurIPS in Vancouver this week🍁 Connect with our #AI #ML researchers @VictorSabanza & @shreyaspadhy. Our research paper on Multi-Fidelity Bayesian Optimization (MFBO) will be @ AIDrugX workshop on Dec 15! Full article: https://t.co/LmBvPgaPyl @AtinaryTech #SDLabs

1/ Starting material constrained synthesis planning is now possible using a general retrosynthesis algorithm *without* training a dedicated value network! Check-out our new paper, TangoStar. Preprint : https://t.co/hVG1r3f6fa

📜 two very interesting, fundamental papers that shed light on when multi-fidelity (MF) Bayesian optimization (BayesOpt) is fruitful, compared to single-fidelity (SF) BayesOpt. https://t.co/qGnWzgYoTt (led by @loic_roch, @pschwllr) https://t.co/lX9VEUyTam (led by @JelfsChem)

5/ This work is the result of my internship at @AtinaryTech , a collaboration between academia and industry working in an amazing team 🧑‍🎓 🤝🧑‍🔧. With the support of @NCCR_Catalysis and @EPFL_ReO GlobaLeaders program! 💪

4/ If you want to see the results in detail, check the preprint: 📄: https://t.co/PFunVQva9H Our paper has also been accepted in the NeurIPS2024 AIDrugX workshop. 🧬💻

3/ ⚗️ We applied our MFBO guidelines to real-world chemistry and materials benchmarks, where a cheap and informative approximation of the problem is available. The result? MFBO consistently reduced costs and improved outcomes compared to standard BO methods.

2/ 🔍 Our study systematically explores the performance of MFBO across different scenarios, revealing that its success depends on the balance between cost and informativeness of low-fidelity data. When low-fidelity data is both inexpensive and informative, MFBO offers advantages.

1/ 🧪💡 Curious about faster material & molecular discoveries? Multi-fidelity Bayesian Optimization (MFBO) is your friend! In this paper, we investigate when MFBO is truly effective compared to standard BO methods, helping to balance cost & accuracy in optimization campaigns.

Just two weeks before we start Phase II of @NCCR_Catalysis! 🎉😍 Earlier this year, we gathered our members & supporters at @ZentrumPaulKlee to celebrate this incredible milestone! 🤩 We’re looking forward to another exciting four years 🥳