🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

🚀 Excited to release a major update to the Boltz-1 model: Boltz-1x! Boltz-1x introduces inference-time steering for much higher physical quality, CUDA kernels for faster, more memory-efficient inference and training, and more! 🔥🧵

BindCraft v3 shows 46 % hit-rate across 12 hard targets. Fresh results add Cas9/CbAgo inhibition, allergen IgE blockade, and AAV capsid retargeting --- no screens, single GPU. Is back-prop-through-AlphaFold now the go-to for binder design?

Knowledge Distillation for Molecular Property Prediction: A Scalability Analysis Distillation does not only work for LLMs, but also for activity/property prediction models for molecules. P: advanced.onlinelibrary.wiley.com/doi/full/10.10…

1/4) I am very happy to share our latest work on the information theory of generative diffusion: "Entropic Time Schedulers for Generative Diffusion Models" We find that the conditional entropy offers a natural data-dependent notion of time.

Limits of deep-learning-based RNA prediction methods 1. This study presents a large-scale benchmark of recent deep learning models for RNA structure prediction, evaluating both single-chain RNAs and RNA complexes to identify their strengths and limitations. 2. Among the nine

AlphaFold’s confidence scores aren’t just QC, they’re latent energy maps 🗺️ Flip pAE/pTM logits into pAEnergy & pTMEnergy and you outscore ipTM, DSMBind, FoldX on SKEMPI, aptamer & mini-protein tests with no labels needed. Do all structure models hide a free docking engine? 😂

Protein structure enablement just leveled up 🚀 Protein loops at 1.1 Å RMSD in 0.10 s per loop ⚡️Lightning-fast⚡️specialist micro-models for hard problems are underrated in this space imo

Co-folding AI will kill docking, but not yet. 🔬AstraZeneca benchmark shows Boltz-1, NeuralPlexer & RoseTTAFold hitting ~3 Å RMSD on orthosteric sites, outclassing classic docking. 🧩 Training bias still places allosterics into the orthosteric pocket, but physics-aware scoring

🧵5 Top Free Alternatives to BioRender for Scientific Illustrations! These five websites offer free scientific illustrations for biologists. Great for presentations, research papers and other research communication needs. Save and share the post!

The first experimentally-solved, 3D structures of the human sweetness receptor. Resolved in complex with sucralose and aspartame; but the structures were identical. Upon binding, these molecules cause the protein to *slightly* bend & trigger a signaling cascade in the cell.

The Open Molecules 2025 (OMol25) Dataset, Evaluations, and Models １．OMol25 is the largest open DFT dataset to date, with over 100 million ωB97M-V/def2-TZVPD single-point calculations covering 83 elements and a wide range of charges, spins, solvation states, and reactivity

The Open Molecules 2025 (OMol25) Dataset, Evaluations, and Models １．OMol25 is the largest open DFT dataset to date, with over 100 million ωB97M-V/def2-TZVPD single-point calculations covering 83 elements and a wide range of charges, spins, solvation states, and reactivity

You can now design binders with 49 different amino acids of both linear and cyclic topologies. We also made it a lot more efficient, for the validated target we only used 1 GPU vs 24 with our previous model. Try it: github.com/patrickbryant1…

Excited to unveil Boltz-2, our new model capable not only of predicting structures but also binding affinities! Boltz-2 is the first AI model to approach the performance of FEP simulations while being more than 1000x faster! All open-sourced under MIT license! A thread… 🤗🚀

Paper: Data Scaling and Generalization Insights for Medicinal Chemistry Deep Learning Models chemrxiv.org/engage/chemrxi…

I'M BLOWN AWAY. Andrej Karpathy just explained Software 3.0 at YC. BIG IDEAS: English is coding. AI is electricity. And, build for LLMs, not just people. Key takeaways:

Atomic resolution ensembles of intrinsically disordered and multi-domain proteins with Alphafold １．A new method called bAIes combines AlphaFold2 with a Bayesian framework to generate atomic-resolution ensembles of intrinsically disordered proteins (IDPs). It uses AlphaFold’s

🧬 AlphaFold DB is aging, as all DB do. 575 / 20,504 human proteins now mismatch UniProt, while zebrafish discrepancies are at 43% 😅 Static structure banks drift as sequences evolve, risking wrong drug‑design templates

CASP is getting cut by NIH... 😢 (Anyone with extra funds wanna help support perhaps the most important competition of the century?) science.org/content/articl…

Zero-shot antibody design in a 24-well plate Chai Discovery 1. Researchers have introduced Chai-2, a multimodal generative model that marks a significant leap in de novo antibody and miniprotein design. This platform achieves an impressive 16% hit rate in de novo antibody