Generative AI in Drug Development

Using #generativeai, we have the ability to customize the molecular structure of antibody drugs, tailoring them to possess ideal properties. It's like engineering a car to be faster, safer, and more fuel-efficient! With this technology, we can enhance the potency of these drugs at lower doses, transforming them into highly effective medicines. The result? Fewer injections for patients, significantly improving both drug safety and reliability. In this recent paper (https://lnkd.in/eiXjqSbm), the authors used a zero-shot approach to create and screen over 1 million antibody variants. Their aim was to design all CDRs in the heavy chain of the antibody specifically for binding to human epidermal growth factor receptor 2 (HER2). The outcome? They discovered three antibodies that bind to HER2 even tighter than trastuzumab(!), along with an additional 23 antibodies that exhibit moderate affinity to HER2. What's even more impressive is that they generously made the sequences open source. I am VERY excited about how this type of research can advance antibody drug development and create a future where personalized and effective treatments are accessible to all.

GenAI-enabled insilico drug design is super hot: just last week we had the $1B funding announcement of Xaira Therapeutics; this week Google DeepMind/Isomorphic Labs (not clear where that boundary is) announced that AlphaFold, which pretty much solved the protein folding problem, has been massively extended to model not just proteins, but also DNA, RNA, and ligands - and the interactions between all those molecular structures. Interesting that the primary enabler of this seems to be the addition of diffusion AI models (that in other applications generate images). These models start with a cloud of digital noise and then incrementally "denoise" that clouds to create molecular structures. (More and more at Ryght we're seeing the utility of repurposing LLMs created for one application in other applications.) Sure feels like Xaira Therapeutics and Isomorphic Labs are on a collision course - both now have GenAI platforms that can (perhaps!) model the biochemistry required for insilico drug design. Here's a recent description of what's coming out of the David Baker's lab, which is the tech engine for Xairia: “Proteins don’t function in isolation,” Baker told Endpoints News. “The fact that AlphaFold and RoseTTAFold only predicted the structure of the protein but not the rest of the system was a limitation.” "The new versions allow researchers to add other biomolecules to the mix, including DNA, RNA, metabolites, drugs and more. Baker’s study, published Thursday in Science, showed how the system could be used to design proteins that bind the oxygen-holding molecule heme and the heart disease drug digoxigenin."

A breakthrough paper in AI for drug discovery from one of our portfolio companies! For the first time, the team of Insilico Medicine “pulled back the curtain” on their AI-driven drug discovery process. In a nutshell, they conducted a generative AI-enabled design of a TNIK inhibitor that targets fibrosis. The molecule has successfully passed preclinical trials and is now in clinical. The results are published in Nature Biotechnology (link in the comments) There are several reasons why this paper is a must-read: - The authors reported on the use of AI to discover and prioritize a new anti-fibrotic target - TNIK - in a separate paper, they reported a "hallmarks of aging assessment" of TNIK to see if the target is relevant for longevity research. It scored high in several hallmarks! - they used Insilico's generative AI for chemistry, Chemistry42 to design a novel molecule with the desired properties. The goal was to maximize safety and efficacy,  among other things. - they performed a large number of in vitro and in vivo experiments, nominating three preclinical candidates for lung and kidney fibrosis - they performed a clinical safety trial in humans -  they also made it interactive via a specialized chatbot, so you can now ask the paper questions, "chat with the paper" (try it yourself, link in the comments).   - they created a substantial data repository with the raw data from multiple experiments. I was personally one of the early investors in Insilico Medicine. A decade later, I am glad the company has become a success story with big ambitions for the future. Already now, the company nominated 17 preclinical candidates since 2021 and five are progressing to clinical stages. Insilico secured major global out-licensing agreements, including partnerships with Exelixis and Menarini in 2023. The paper is a must-read if you want to know how to apply AI to discover novel drugs!  Alex Zhavoronkov Alex Aliper Jan Szollos, MBA #artificialintelligence #drugdiscovery Image credit: Insilico Medicine

Google creates AI Large Language Model (Tx-LLM) for Drug Discovery and Therapeutic Development: 💊Tx-LLM uses PaLM-2, Google’s generative AI technology that uses their LLMs to answer medical questions 💊Tx-LLM can handle a wide range of chemical or biological entities, including small and large molecules, proteins, and disease targets and demonstrates strong knowledge transfer between various drug types due to the diversity of its training datasets 💊The LLM combines free-text instructions with representations of small molecules, such as SMILES strings for small molecules 💊 SMILES, or Simplified Molecular Input Line Entry System, is a text input using printable characters that represent molecules and chemical reactions 💊 To predict how well drugs work together, researchers used prompts that included instructions, background information, and a specific question. 💊 The model can be asked to predict things such as whether a particular molecule can cross the blood-brain barrier 💊 The authors noted that training on a variety of datasets, including biological sequences, actually helped improve performance on tasks involving different types of drug data. 💊 Unlike models that specialize in single tasks, Tx-LLM integrates knowledge across various stages of the drug development process e.g. early-stage target discovery to late-stage clinical trial approval, thus enhancing contextual understanding and overall performance. 💊 Future enhancements could include integrating additional models, such as the Gemini family, and improving the model's ability to explain its predictions 👇 Link to article and paper in comments #DigitalHealth #AI

My latest: VantAI just announced a deal with Bristol Myers Squibb potentially worth up to $674 million. The goal? To use generative AI to design drugs that use the body's own cellular waste disposal process to fight disease. The key is using a combination of LLMs and the same graph theory algorithms that TikTok uses to send you Travis and Taylor content to design what are called "molecular glues." These glues bind to pathogens in the body, like cancer cells, connecting them to the protein degraders your cell uses to break down stuff it doesn't need anymore. Why AI? Because your body doesn't naturally use this process to fight disease, VantAI CEO Zach Carpenter explained to me. "You can't just discover glues through trial and error." You have to design the molecules that do it, a Herculean task without the last decade's advances in machine learning. Read more.