Leonard Wossnig’s Post

AI is opening up a new era in biomolecular research, Google DeepMind, along with Isomorphic Labs, created AlphaFold 3. The AI tool extends its predictive capabilities to not just proteins but also DNA and RNA, opening new avenues for research and pharmaceutical development. The model can interact between different types of biomolecules, which is crucial for understanding complex biological processes and accelerating drug discovery. With the AlphaFold Server now freely available, researchers from all corners of the globe can harness the power of superior molecular predictions, fostering a truly global and collaborative approach to biomolecular research. While we enjoy AI's daily conveniences, we must recognize cases where tech can bring together scientific communities to advance research in a certain area. #medtech #google #biotechnology

📰 News: 📰 🔬🚀 Exciting News! OpenFold AI Research Consortium Welcomes Six New Members! 🔬🚀 OpenFold, the leading AI research consortium dedicated to advancing biology and drug discovery, is thrilled to announce the addition of six new industry members. 🌟 Joining the consortium are Astex Pharmaceuticals, Biogen, Congruence Therapeutics, Polaris Quantum Biotech, Psivant Therapeutics, and SandboxAQ. 🎉 💡 These companies bring unique expertise and a strong commitment to harnessing the power of AI to accelerate drug discovery and ultimately improve patient outcomes. Through collaboration and innovation, they will contribute to OpenFold's groundbreaking AI platform development programs. 🌍 This expansion signifies a significant leap forward in AI-driven solutions for healthcare and life sciences. By joining forces, OpenFold and its members aim to revolutionize medicine by developing transformative drugs and leveraging cutting-edge technologies. 🧪💊 ⚛️ With the inclusion of Polaris Quantum Biotech, we are also excited to explore the potential of quantum computing in the field of drug discovery. Quantum computing has the potential to revolutionize the way we tackle complex problems in various industries, including healthcare. 💡 💼 Interested in exploring quantum solutions for your small and medium-sized business -SMB-? This could be the differential factor that boosts your efficiency and propels you ahead of the competition! Don't miss out on this incredible opportunity. 👍 Join the conversation and share your thoughts in the comments below! Let's discover the limitless potential of quantum computing together. Follow Spin Quantum Tech for more exciting updates and insights. 🌐📲 -AI -QuantumComputing -Innovation -BusinessGrowth -SpinQuantumTech #Future #Futurism #Innovation #Business #ArtificialIntelligence #QuantumComputing #SQT https://lnkd.in/eE2p2URH

AlphaFold3 (AF3) is making waves in the world of drug discovery, but its true potential—and limitations—are still being uncovered. Leonard Wossnig (CTO of LabGenius) co-authored a deep dive, with Alexander Mathiasen (Senior Machine Learning Engineer at Charm Therapeutics) that explores the impact of AF3 on computational drug discovery - it’s a must-read for anyone in the field! One key issue they address? Data scarcity. AF3 was trained on the Protein Data Bank (PDB)—however, with only that limited data available models like AF3 plateau on accuracy gains. As Wossnig highlights, improving these models requires new, innovative data generation methods, possibly leveraging physics-based approaches or molecular dynamics simulations. 💡 Another alternative worth considering is accessing and securely collaborating on proprietary data across organizational boundaries, which can provide richer datasets and help overcome the limitations of internal or public data alone. 🌐 For a deeper look at AF3’s strengths, challenges, and where the field needs to go next, check out the full article (shared as comment) #AlphaFold3 #AIinBiotech #DrugDiscovery #BiotechInnovation #ML

AI advances in early drug pipelines. Join Recursion experts Julie Owen and Noel Powell at Discovery US 2024 in Boston Nov. 7-8, an event showcasing the key platforms and technologies impacting early drug pipelines, from DNA-encoded libraries and phenotypic screening to AI algorithms and organoid models. ◾ On Nov. 8, 10:20am on Track 1, Julie will present  “Targeting The Unknown: AI’s Shortcut To Drug Discovery,” explaining how we are using AI and data-driven approaches to radically improve the efficiency of drug discovery – commanding massive experimental and computational scale and uniting technology, biology and chemistry to advance the future of medicine. ◾ On Nov. 8. 9:30am on Track 2, Noel will present “Recursion’s Phenomic Map-Based Drug Discovery Approach: From Project Ideation To Lead Optimization,” discussing how the Recursion Operating System provides a mapping and navigating approach that enables us not only to unravel the complexity of biology but also to identify chemical starting points and drive SAR, allowing us to efficiently advance projects from initiation through different stages of pre-clinical development. 👉 Learn more and register here: https://lnkd.in/eiVG3w28 #OGDiscoveryUS24 #DiscoveryUS #DiscoveryUS24 #chemistry #smallmolecules #drugdiscovery #techbio #AI

Introducing DiffLinker: A Breakthrough 3D Atomic Point Cloud Model 🔗 DiffLinker aims to generate molecular linkers between input fragments using a neural network. It determines and iteratively updates the size, initial types, and positions of linker atoms to create a connected molecule. Key Advantages of DiffLinker: ➡️ Generates linkers without preset size limitations ➡️ Equivariant to translations, rotations, reflections, and permutations ➡️ Not restricted by the number of input fragments In practical drug design applications, DiffLinker demonstrates remarkable effectiveness in producing chemically meaningful linkers, connecting multiple fragments while considering geometric constraints imposed by nearby protein atoms. The article showcases DiffLinker's utility in various scenarios, including enhancing JNK selectivity and designing inhibitors for Hsp90 and IMPDH. In summary, DiffLinker represents a promising tool for molecular linker design in drug discovery, facilitating the generation of synthetically accessible and drug-like compounds. #DrugDesign #Chemistry #Innovation #DiffLinker #Molecules #NeuralNetwork #Fragments #Research https://lnkd.in/e8PSDvuZ

Maruan Al-Shedivat, PhD and the entire ML research team at Genesis Therapeutics are doing some of the most compelling work right now in AI. Seriously exciting and important work. Watch Maruan's recent talk at ICML on drug discovery with generative modeling.

🎉 Glad to share our breakthrough in molecular glue-mediated ternary structure prediction! By introducing an enhanced sampling inductive bias into the diffusion module in the Alphafold3-type models, we achieved remarkable results, enabling inference scaling laws to emerge. This breakthrough powers YDS-Ternoplex, which successfully predicts all five novel ternary interfaces (not included in the training data). The most exciting validation came from the FKBP12:MG:mTOR-FRB complex prediction. Despite having an alternative protein-protein interface in the training data that could have led models to memorize and reproduce known interactions, YDS-Ternoplex correctly predicted the novel interface configuration. This strongly suggests our model’s generalization capability. A key insight I would like to share here is, the success of this inference-only enhancement implies that the Pairformer architecture has effectively captured the biomolecular interaction landscape. This opens exciting possibilities for future applications in structure-based drug design. The predicted structures are in SI. #DrugDiscovery #AI #StructuralBiology #MolecularGlues #Innovation #Alphafold3 Happy Holidays. Happy New Year. 🎉 🎊 https://lnkd.in/gXG8e5TQ

Alphafold server by Google, a wonderful application powered by Deepmind's AlphaFold 3. Previously we needed 1 PhD student study (3-4 Years) to get the structure of a molecular protein fully. But now with AlphaFold, we can successfully predict the structure of molecular proteins with simple GATC spelling. How does it impact humanity? We can speed up the development of drugs, simulation of drug-drug interactions and even make our own digital twins out. This technology is now free and open source to every single biologist in the world, I can't wait to see how this innovative AI model will disrupt the pharmaceutical industry. Try it here yourself: https://lnkd.in/gD5gXrNU

I'm thrilled to share that our paper, ‘𝘋𝘪𝘨𝘪𝘓𝘰𝘊𝘚: 𝘈 𝘓𝘦𝘢𝘱 𝘍𝘰𝘳𝘸𝘢𝘳𝘥 𝘪𝘯 𝘗𝘳𝘦𝘥𝘪𝘤𝘵𝘪𝘷𝘦 𝘖𝘳𝘨𝘢𝘯-𝘰𝘯-𝘊𝘩𝘪𝘱 𝘚𝘪𝘮𝘶𝘭𝘢𝘵𝘪𝘰𝘯𝘴’, has been accepted for publication in PLOS ONE!🎉 𝗪𝗵𝗮𝘁'𝘀 𝗶𝘁 𝗔𝗯𝗼𝘂𝘁? We developed DigiLoCS, a digital liver-on-chip simulator for modelling the clearance in a liver-on-chip system closely mimicking the human liver clearance. Our approach uses a compartmental physiological model of the liver, based on ordinary differential equations (ODEs), to estimate pharmacokinetic (PK) parameters related to liver clearance. 𝗧𝗵𝗲 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲 Accurately predicting 𝘪𝘯 𝘷𝘪𝘷𝘰 clearance from 𝘪𝘯 𝘷𝘪𝘵𝘳𝘰 data is important as inadequate understanding of the clearance of a compound can lead to unexpected or undesirable outcomes in clinical trials, ranging from underdosing to toxicity. 𝗢𝘂𝗿 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻 DigiLoCs facilitate accurate description of on-chip biology to disentangle biological processes, namely clearance, permeability, and partitioning. The model uses ODEs to define the drug concentrations in media, interstitium and intracellular compartments based on biological, hardware, and physicochemical information. 𝗞𝗲𝘆 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆𝘀 • Predicting human 𝘪𝘯 𝘷𝘪𝘷𝘰 clearance across different hepatic 𝘪𝘯 𝘷𝘪𝘵𝘳𝘰 systems. • A digital twin framework bridging the gap between 𝘪𝘯 𝘷𝘪𝘵𝘳𝘰 findings and clinical relevance. 𝗧𝗵𝗲 𝗜𝗺𝗽𝗮𝗰𝘁 DigiLoCs can serve as a powerful decision-support tool in pharmaceutical research, helping estimate first-in-human doses, evaluate human pharmacokinetics, and reduce the need for animal testing. This approach is also adaptable to other physiological contexts, extending beyond liver metabolism to models of the gut, brain, or placenta. Big thanks to my co-authors, Christian Maass, Chitta Mandal, Alex Pothen, and Stephan Schaller, for their collaboration—and to our peer reviewers for their insightful feedback. Looking forward to next steps in advancing organ-on-chip modelling! 🔗 Link to preprint https://lnkd.in/e6ECCwpu #PaperAcceptance #Research #Computationalmodelling #OrganOnChip

Quality content. Flexible data models. Power your LLM with the content you already trust. Elsevier journals. The overall innovation picture for you and Elsevier? see Mirit’s post below! 💡