Valence Labs’ Post

Google DeepMind's new AI model, AlphaFold 3, can predict the structure of not just proteins but also DNA, RNA, and other molecules, improving 50% on previous models. This expansion into modeling life's molecules will aid researchers in various fields, including medicine and drug development. The model, which uses a diffusion method to generate 3D models of new structures, has been used by DeepMind's drug discovery company, Isomorphic Labs, to improve understanding of new disease targets. A free AlphaFold Server, powered by AlphaFold 3, is available to some researchers for generating biomolecular structure predictions. Google is working with the scientific community and policy leaders to deploy AlphaFold 3 responsibly, addressing potential biosecurity concerns. #google #deepmind #dna #technology #technews #tech #news #techupdates

This is a great interview that covers a lot of ground. A key thing in it is something here at Leash we call "cultural arbitrage". The unique culture Chris Gibson and Blake Borgeson built at Recursion was made by combining wet (experimental) and dry (computational) people in the same room and getting them to listen to each other. With that crew, Recursion has constructed what is probably the largest ML-bespoke biological dataset on earth. Such a culture is difficult to replicate! You need wet people who are willing to add the extra controls the dry people want. You need dry people willing to be patient with the huge amount of effort the wet people make to produce the data in the first place. You need to educate both groups on what can and cannot be done. You need everyone talking constantly to increase the speed of your experiment/analysis iteration cycles. I've seen wet groups try to bolt on a dry group or vice-versa and it rarely goes well. Almost never, really. Smoothly-running teams have a distinct, underappreciated advantage. They're able to generate better data faster than the other teams. Having such a team is cultural arbitrage. How can you get your teams to get better at this? Over here, we get the data engineers to pipet and the bench scientists write SQL queries. Everyone takes extra time to teach each other critical details instead of simply generating data and throwing it over a wall to the next person. We like sprint planning and daily standups (from software culture). We also like journal clubs and narrative note-taking (from lab culture). We build webapps so that everyone in the company can look at any, or all, experimental results immediately and see if they pass the smell test. Crucially, the dry people help design experiments (more replicates! more controls! more randomization!) and the wet people communicate sources of variation (reagent lots! different donors!) and pass them on as metadata for models to ingest. This stuff isn't particularly glamorous but in our experience is absolutely necessary for ML to make a strong dent in biological problems. We believe cultural arbitrage is one of the most important things to cultivate as soon as you can. Go do it!

🚀🔬 TikTok’s Parent Company Is Moving Into Pharma Based on this article by Forbes, the Beijing-based tech giant is recruiting American talent in computational biology, quantum chemistry, molecular dynamics, and physics for its “AI for Drug Design” and “AI for Science” teams. An insightful take on the potential for data and algorithms combined with open AI to be a bridge to Drug Discovery and computational biology. Regardless of the potential, healthcare offers opportunities for diversification and focus into fields viewed more positively by governments and regulators. We have seen this with Goldman Sach's $650 million fund for life sciences investments if it is any indication of what 2024 has in stock. #aidrugdiscovery #lifesciencesindustry #healthtech

Exciting news at CytoReason! As a Senior Bioinformatics Scientist working for Science Operations at CytoReason, I am thrilled to announce that we have secured our Series B funding. We have received significant backing from NVIDIA, Pfizer, and Thermo Fisher Scientific, as well as continued support from OurCrowd and Asymmetry Ventures. ❗ Why is this important? 📱 Technology Boost: Our platforms have been powered up using NVIDIA's latest #AI technology, which has resulted in over 10x faster inference workloads. This upgraded technology will enable more life sciences companies to utilize our predictive clinical insights. 💻 Data Power: By integrating Thermo Fisher's extensive data with our AI, we are poised to revolutionize pharma R&D and advance personalized healthcare. ↗ Future Potential: Our collaboration with Pfizer will enhance their R&D capabilities, creating new opportunities for drug development and improving patient outcomes. You can learn more about this exciting news at: https://lnkd.in/dU5YrCYp This is a proud moment for CytoReason and the industry, and I am excited to be part of this journey.

Google DeepMind and Isomorphic Labs have unveiled AlphaFold 3, a groundbreaking AI model that predicts the structure and interactions of all life's molecules with unprecedented accuracy. AlphaFold 3's advancements, showcased in Nature, offer a profound leap in understanding biological processes and drug discovery. It surpasses existing methods by up to 50%, doubling accuracy in key interactions. Its free-to-access AlphaFold Server democratizes access to cutting-edge research tools, empowering scientists worldwide. Moreover, AlphaFold 3 revolutionizes drug design, partnering with pharmaceutical companies to develop life-changing treatments. This monumental achievement builds upon AlphaFold 2's success and promises transformative science, from enhancing crop resilience to accelerating genomics research and drug design. . . . #google #ai #artificialintelligence #science #innovation #technology #future

Amazing breakthroughs are being made at Prescient Design, with our Director of Frontier Research, Stephen Ra, leading the charge in executing a paradigm shift in drug discovery. Our Frontier Research group is dedicated to exploring fundamental concepts, algorithms, and theories in machine learning, and using them to enable scientific discovery. We're excited to be developing methods in adaptive and controllable generation in generative models, uncertainty quantification, robust representation learning, and optimization of the interface between in silico frameworks and in vitro/in vivo experiments. Learn more about our game-changing Computational Sciences (gCS) organization at Genentech here: https://lnkd.in/gBm9gq3T #artificialintelligence #machinelearning #deeplearning #generatieveai #drugdiscovery #medicine #weareroche

🌟𝙊𝙥𝙚𝙣-𝙨𝙤𝙪𝙧𝙘𝙚 𝙨𝙘𝙞𝙚𝙣𝙘𝙚 𝙨𝙘𝙤𝙧𝙚𝙨 𝙖 𝙬𝙞𝙣 𝙞𝙣 𝙙𝙧𝙪𝙜 𝙙𝙞𝙨𝙘𝙤𝙫𝙚𝙧𝙮!💊 #BioNeMo #GTC24 Yesterday, the CEO of NVIDIA , Jensen Huang, gave a shoutout to the power of open-source science in the field of drug discovery. 🙌 One of the highlights was the inclusion of DiffDock in BioNeMo, a platform that helps scientists develop new drugs faster and more efficiently. DiffDock is an open-source software tool created by researchers at the Jameel Clinic. It uses AI to predict how small molecules bind to proteins, which is a crucial step in drug discovery. By making DiffDock open-source, the MIT Jameel Clinic - AI & Health team has made it possible for other scientists around the world to use and improve upon their work. This kind of collaboration is essential for accelerating progress in drug discovery. So, a big round of applause to the researchers involved in this project! 👏 Their work is a shining example of how open-source science can make a real difference in the world. Keep up the great work, team! 💪 #OpenScience #DrugDiscoveryRevolution

A Superbio Scientist Spotlight of Petrina Kamya, Ph.D.! The team at Superbio.ai had a terrific interview with Petrina Kamya, Ph.D., Insilico's Global Head of AI Platforms and President of Insilico Medicine Canada about her work and driving interests -- from her training in quantum chemistry, to her early interest in AI drug discovery, to the ways in which AI is changing the paradigm for drug discovery at every stage of the process. ◾ On quantum drug discovery: "Now, with the advance of technology, quantum chemistry is coming back, and with the development of quantum computing as well, it’s beginning to make inroads in drug discovery and drug development that could be very impactful." ◾ On the difference between AI models: "When you’re using generative AI versus predictive models, those are very different applications of artificial intelligence or machine learning. With a generative algorithm, you don’t need massive amounts of data to train it." ◾ On the role of LLMs: "Because it’s a large language model, it can take in information from omics data, biology, chemistry, clinical trials, and synthesize it into a format or a digestible way. So, you can ask this large language model or knowledge graph questions within a specific disease area. I think they’re very helpful tools." Thank you Berke Buyukkucak for having us on! *Link to interview & video in comments #ai #science #drugdiscovery #career #biotech

Love this post from Najat Khan, PhD, who has just joined Recursion as Chief R&D Officer, Chief Commercial Officer and Board member. She not only talks about what Recursion is doing to rapidly advance and improve end-to-end drug discovery -- including moving beyond phenomics to multiomics, capturing digital animal data, delivering highly optimized "made in Recursion" molecules and AI solutions for pharma, and rapidly advancing preclinical and clinical programs -- but she also gets into the "how," in particular the "bilingual talent and culture that embraces technologists and life science specialists as equals" that she calls "the secret sauce that allows Recursion to adapt and scale." 👇

A wonderful post about scientists trying to shape the future and bring new treatments to patients. Well done Najat Khan, PhD , Recursion and NVIDIA You will hear more in November at CNS Summit.