Aleksandra Faust

De novo protein binders against undruggable targets? Here's how Institute of Protein Design scientists target Intrinsically Disordered Regions using RFdiffusion Try the protocol on Tamarind Bio now: https://lnkd.in/gbS7P92R The authors design binders for 43 targets (18 synthetic peptides + 21 diverse native IDRs + a few others, including polar targets) yielded binders for 39/43 while testing ~22 designs per target. 34 binders reached 100 pM–100 nM Kd, with most other being in the nanomolar range. Unlike most de novo design publications, we have specificity data! All-by-all binding showed each design strongly binds only its intended target. The authors build a template library of pocketed scaffolds that “wrap” extended IDR conformations; target sequences are threaded and then optimized by ML sequence/backbone refinement. This methodology is a promising step towards systematic design of de novo binders against IDRs, with clear applications in proteomics and cell biology, and potential for therapeutics Congratulations on the Science publication! Institute for Protein Design, University of Washington ————— Tamarind Bio is the leading provider of molecular design tools, such as IDR binder design, RFdiffusion, and AlphaFold. Through our web interface, Copilot, and API, scientists can use these tools at large scales, no coding required.

213

“We can have people constantly inserting new experiments, and the lab’s going to manage everything, calling us from time to time just to clean some things” In awe of the engineering behind Berkeley Lab's “A-Lab," an autonomous facility for accelerating materials discovery. During my tour and conversation with David Milsted, a postdoctoral researcher at the lab, we explored: • How automated labs predict, synthesize, and test materials • The orchestration of autonomous experiments through the A-Lab OS • The future of materials characterization using automated microscopy • Behind-the-scenes insights into running complex autonomous lab systems (robotic dispensers, speed mixers, automated furnaces, and more!) If you’re excited by automation, materials science, and seeing how labs can drive innovative research around the clock, this episode is for you. Link to the full episode in the comments below 👇

34

4 Comments

Our latest edition of Beyond the Lab focuses on Dr. Boubacar Kante, whose research focuses on wave-matter interaction.    📌 One of his lab's biggest accomplishments was solving a six-decade old challenge in wave physics and optics by developing a new laser called the Berkeley Surface Emitting Laser, or BerkSEL.   📝 Advice for young researchers: "I want them to be aware of issues beyond the projects they are working on...It helps to put your projects into context, into the bigger picture, and see how you can contribute to a better environment for all of us." Go Beyond the Lab with Boubacar Kanté ➡️ https://lnkd.in/gJgDpvCK #sciencephilanthropy #sciencefunding #physics #mooreinventorfellows #scientistpov

279

3 Comments

The Wyss has refined its approach to AI and launched the Translational AI Catalyst – an institute-wide effort designed to enable and accelerate innovation through the effective and responsible application of new computational approaches, while leveraging machine learning algorithms to generate unique biological datasets. Led by Core Faculty member Jim Collins, Computational Biologist Megan Sperry, and Business Development Director Ally Chang, PhD, MBA, the Translational AI Catalyst will apply AI to enhance application-driven science and address challenges in healthcare and sustainability. https://lnkd.in/eK7k9TA7

16

PhD student Zixi Li is pushing the boundaries of quantum hardware at Pritzker School of Molecular Engineering at the University of Chicago. Her work uses thin diamond films with atomic-scale defects as controllable qubits. This effort brings us closer to practical quantum technologies, from global quantum networks to real-world quantum devices. “I’m very proud that I’m a part of this second quantum revolution and part of this quantum hub,” she said. Learn more about PhD programs at UChicago PME: https://lnkd.in/gKd4xEhh

76

🚨 Two AI Legends Just Said the Same Thing: LLMs Aren't the Path to True Intelligence Richard Sutton (Father of Reinforcement Learning, 2024 Turing Award Winner) and Yann LeCun (Meta's Chief AI Scientist) are both moving beyond LLMs. Here's why this matters for everyone: 🧵 The Problem They Both See 🤔 Current AI chatbots (like ChatGPT) are basically sophisticated copy-cats. They: Mimic what humans would say ✍️ Don't actually understand consequences Can't truly learn from mistakes Have no real goals beyond predicting text Think of it like this: A parrot 🦜 can repeat "the sky is blue" perfectly, but it doesn't understand what sky, blue, or "is" means. What Real Intelligence Needs 🧠 Both experts say we need AI that: 1️⃣ Learns like a child does 👶 Not by copying, but by trying things and seeing what happens 2️⃣ Understands the physical world 🌍 Can predict: "If I push this cup, it will fall" 3️⃣ Has actual goals 🎯 Not just "predict the next word" but "achieve this specific outcome" 4️⃣ Plans ahead ♟️ Like chess players thinking multiple moves ahead, not just reacting The Future They Envision 🔮 Instead of bigger chatbots, imagine AI that: Learns continuously from experience 📈 Updates its understanding in real-time Can actually reason about cause and effect Builds mental models of how things work Why This Matters to You 💡 We're potentially at a turning point. The current AI boom might hit a wall 🧱, and the next breakthrough will come from completely different approaches. For businesses: Don't just bet on scaling current AI - watch for fundamentally new architectures For careers: Understanding real-world interaction and planning might become more valuable than prompt engineering The Bottom Line 📝 Two of AI's most influential thinkers are saying the same thing: We've been building really impressive calculators 🖩 when we should be building systems that actually understand the world. The next phase of AI won't just be ChatGPT-5 or GPT-6. It might be something entirely different. 🚀 What do you think - are we heading toward a major shift in how we approach AI? source: https://lnkd.in/gsHZvNbE https://lnkd.in/gchsKwsE #AI #GenAI #AGI #LLM #

41

3 Comments

Northwestern Engineering’s Emma Alexander, Aggelos Katsaggelos, Han Liu, and Aravindan Vijayaraghavan are helping lead transformative projects at the Vera C. Rubin Observatory as it prepares to launch the Legacy Survey of Space and Time. Their work uses advanced AI to uncover hidden galaxy structures, model stellar evolution, and optimize real-time observation. Northwestern University Computer Science Department https://lnkd.in/ggE9DPXw

14

The next talk in our deep learning performance tutorial features Amir Gholami from UC Berkeley Electrical Engineering & Computer Sciences (EECS) presenting work to accelerate inference of LLMs with long context length

58

1 Comment

🚀 The largest public catalytic dataset, AQCat25, was released in collaboration with NVIDIA DGX and SandboxAQ. It is publicly available as of today, and it is free for academic research. 🧪 If you work in catalysis, you know the pain: running DFT calculations is slow. Testing a single catalyst candidate can take days or weeks of compute time. With enough high quality data, we can train ML models that predict catalytic behavior in seconds instead of days. We are seeing speedups of 20,000x compared to traditional DFT while maintaining the accuracy you actually need. Most existing datasets cut corners to save compute time. AQCat25 doesn't: ✔️ Broader chemical space ✔️ Transition states included ✔️ Higher accuracy calculations ✔️ Models spin polarization 🧠 90% of all chemicals produced globally depend on catalysis. Better catalysts mean less energy waste, lower emissions, and faster paths to things like green hydrogen, sustainable fuels, and more efficient fertilizer production. Relevant link in the comments #Chemistry #AI #MachineLearning #SmallMoleculeDiscovery #DrugDiscovery Chemetrian

19

1 Comment

For the California fusion innovation ecosystem, the moment is now. This is one of my takeaways from the statewide fusion convening we just held at UC San Diego, in conjunction with General Atomics and Lawrence Livermore National Laboratory. Right now, universities, national labs and federal agencies are working together to solve the remaining R&D challenges that are holding back the ability of fusion startups to deliver power plants capable of running 24/7. Also right now, fusion startups in California and around the country are in direct conversation with local and state government agencies regarding where to build pilot plants and future manufacturing facilities. The fact that both of these things — and so much more — are happening right now across our fusion ecosystem is an indication of the dynamic, fast changing fusion R&D and commercialization environments we are all now operating in. As an engineering dean at a school with a long-standing focus on fusion and fusion-enabling technologies, it’s clear to me that universities and others in the public sector need to meet this moment. And in fact, there was broad consensus at our California fusion convening that we need to accelerate when it comes to public-private coordination across the state. Here at UC San Diego, we are certainly giving it our best effort. Just over a year ago, we launched our UC San Diego Fusion Engineering Institute. Since then, momentum has continued to build. Some recent key achievements:  ▪️ Jacobs School engineers were awarded $4M for a new collaborative fusion research effort funded by the University of California National Laboratory Fees Research Program to address tough shared challenges previously surfaced by the California fusion community.  ▪️the PISCES Lab is now running with an incredible upgrade. A $15 million U.S. Department of Energy award allowed for the addition of an ion beam accelerator to his lab’s existing experimental infrastructure. The upgraded facility fills an important need for fusion researchers in the public and private sectors working to develop new fusion reactor materials. ▪️Earlier this year, scientists and engineers at General Atomics and UC San Diego established a Fusion Data Science and Digital Engineering Center in San Diego. ▪️This year, UC San Diego became the new home to the nonprofit Fusion Power Associates (FPA). This is just scratching the surface — there is much more to come. If you’re interested in empowering UC San Diego to meet the fusion moment, please get in touch. I can’t overstate the value of graduate fellowships in fusion — to give just one example. Read more about all of these fusion updates in our Jacobs School November newsletter: https://lnkd.in/gE2ZCvUx In the meantime, wishing each of you a wonderful Thanksgiving and restful weekend. 

62

2 Comments

I'm thrilled to share that our new paper, "Lyapunov-Based Physics-Informed Deep Neural Networks with Skew Symmetry Considerations," is officially out on arXiv! This research is focused on leveraging the known structure and properties of the Euler-Lagrange equation within a deep learning framework while ensuring stability. Our key innovation is embedding physical knowledge into the Lyapunov-based Deep Neural Network's (Lb-DNN) adaptation process via a skew-symmetric prediction error which penalizes updates that violate known system properties. A huge thank you my co-authors Ciku Makumi, PhD, Rushikesh Kamalapurkar, and Warren Dixon for their work and collaboration on this project. Read the Full Paper Here: https://lnkd.in/gX-VHRjS

155

2 Comments

We released a big cat on Hugging Face today! 😸 😸 😸 AQCat25 is the first large-scale catalytic dataset including magnetic effects spanning a diverse chemical space. This is SandboxAQ's second large-scale dataset, made publicly available to accelerate AI-driven scientific discovery. AQCat25 in numbers: 🏅 11 Million data points 🧪 40K Intermediate catalyst systems 🚀 400K GPU-hours of compute to generate Catalysts power 90% of chemicals and 80% of all manufactured goods globally. AQCat25 is a dataset to revolutionize how we discover and optimize them with AI. Developed on NVIDIA DGX™ Cloud, AQCat25 enables highly accurate predictions of material properties from atomic structures, opening new possibilities across energy, chemicals, automotive, agriculture, and consumer goods industries. If you want to learn more about how AQCat25 is accelerating AI-driven catalyst design, sign up for our upcoming webinar on the webpage. To learn more: Visit AQCat25 webpage: https://lnkd.in/gVuDDNVY Read our blog: https://lnkd.in/gn7rshxZ Read the Press Release: https://lnkd.in/gBvypXBX Download the data on HF: https://lnkd.in/gdghY-J6 #AQCat25 #SandboxAQ #AIforScience #Catalyst #MaterialsDesign

110

9 Comments

Prof G / Gil Luria nailed the NVIDIA point on yesterday's earnings: the real bottleneck isn’t silicon, it’s infrastructure. The same theme is true in precision medicine: breakthrough assays or NGS platforms can only scale if reimbursement, regulation, and clinical adoption are ready. Just like Nvidia needs power and HVAC, diagnostics need the “plumbing” of healthcare systems to turn innovation into outcomes. Here’s the paradox: NVIDIA is the most valuable company in the world ($4T!!) and the markets recognize the infrastructure AI requires. Yet in medicine, the companies building the engines of health (Dx, NGS, precision tools) remain deeply undervalued, even though they determine outcomes for billions. We’ve rewarded compute far more than care. Maybe we just need to start calling Dx “GPUs for the body”! ;) Link in the comments, for those interested!

129

26 Comments