Gleb Kuznetsov

The Silent Tutors: How LLMs Are Bridging Knowledge Gaps in Cell Therapy Manufacturing

3

A new study from Westlake University introduces #Pinal, a groundbreaking approach to de novo protein design using natural language guidance. 🎯 This new tool generates protein structures from language inputs and then designs sequences based on both structure and instructions. It outperforms existing models in creating diverse, functional proteins from scratch. This two-stage method promises to expand the frontiers of protein engineering. Quick Read: https://lnkd.in/g4CMkDzH #bioinformatics #proteindesign #proteinlanguagemodels #LLM #AI #syntheticbiology #proteinengineering #drugdiscovery #sciencenews

47

1 Comment

A lot of insightful articles have already been written around the recently announced Recursion/Exscientia merger in the AI drug discovery space, including the great comments from In-Silico's Alex Zhavoronkov, and more recently from STAT's Allison DeAngelis. Both posts are very much worth checking out (STAT is behind a paywall unfortunately). I'm now very curious to see how the AI drug discovery space will evolve. Will it become more of a dedicated SaaS or CRO type-play to other companies, or will we revert back to the "traditional" biotech drug-development model all the way to phase II, just with AI enablement? I personally would be willing to rather bet on the latter, as most of the molecules will eventually be picked up by big pharma anyhow, and the valuation model for compounds is so much more established than for AI tools. This, if you believe the STAT article, is clearly the route that Recursion will be taking, after, as nicely reported by STAT, the earlier stage discovery partnerships with pharma are still somewhat tricky. Furthermore, until we truly have foundational biology models (which depending on who you ask is still 5-10 years away), the role of the wet lab and traditional chemistry and large molecule development skills shouldn't be underestimated, AI still only gets you so far (though it can help optimize many of the steps in the journey). Would love to hear your thoughts on this topic. https://lnkd.in/e9asK4-d https://lnkd.in/etfHC2v2

12

The Era of 1-bit LLMs “Every single parameter (or weight) of the LLM is ternary {-1, 0, 1}. It matches the full-precision (i.e., FP16 or BF16) Transformer LLM with the same model size and training tokens in terms of both perplexity and end-task performance, while being significantly more cost-effective in terms of latency, memory, throughput, and energy consumption.

7

A couple of announcements for ISoP Members, 1. Today (August 5th) is the last day for early bird ACoP 2024 conference registration. Please see link below to register! https://lnkd.in/eeYTx_WY 2. ISoP QSP SIG Virtual QSP Week is next week (August 12-15). My own research interests certainly bias me here, but I think this will be the best Virtual QSP Week yet! There are several talks discussing how AI/ML can be leveraged for identifying biomarkers, clustering patient populations to understand treatment response, and the use of generative AI and LLMs in QSP model development and applications. Don't forget to register and add these events to your calendar in advance (instead of trying to find the zoom link day of). The way it works is that you need to click the registration links below for each event, you will get an email, and then you need to click add the event to your calendar for each of the 4 separate webinars. Please see more information on each of webinars below. Agenda and Speakers: Session 1. 12 Aug 2024, Monday, 9 am EDT: Registration Link: https://lnkd.in/er9jeTXM AI-driven exploration of dementia biomarkers Chairs: Suruchi Bakshi, Shaina Short Presenter: Dea Gosishvili, University of Amsterdam Session 2. 13 Aug 2024, Tuesday, 9 am EDT: Registration Link: https://lnkd.in/eaxqPyf5 From molecules to systems: leveraging NEURON for multiscale modeling      Chairs: Suruchi Bakshi, Shaina Short Presenter: Robert McDougal, Yale University   Session 3. 14 Aug 2024, Wednesday 11 am EDT Registration Link: https://lnkd.in/ew2ZqusU Using multi-omics clustering and machine learning to identify key characteristics driving treatment response in SLE virtual patients       Chairs: Prakash Packrisamy, Vincent Hurez Presenter: Renee Myers, Rosa & Co. LLC Application of GenAI in QSP model development       Chairs: Prakash Packrisamy, Vincent Hurez Presenters: Anuraag Saini and Ali Farnoud, Boehringer Ingelheim Session 4. 15 Aug 2024, Thursday11 am EDT Registration Link: https://lnkd.in/ekKGdZht Large Language Model tools for QSP modeling       Chair: Carolyn Cho Presenters: Chris Rackaukas and Emily Nieves, Massachusetts Institute of Technology Link to QSP SIG Website and Virtual QSP Week: https://lnkd.in/e96SdAsh Lastly, if you have not yet checked out the new ISoP Members Community please do so. Trying to build a vibrant environment for the community to engage in conversation is extremely difficult. We are trying to figure out the features of this new application, your help is appreciated throughout this process! Link below: https://lnkd.in/eBZWrscu

5

Scientists unveil #GeneRAG: A new framework enhancing LLMs' gene analysis capabilities. GeneRAG outperforms GPT-3.5 and GPT-4 in gene-related tasks, showcasing the power of Retrieval-Augmented Generation in genomics. Quick Read: https://lnkd.in/gyiyM28F #bioinformatics #LLMs #AIinBiology #datascience #ai #sciencenews

8

He used to be a postdoctoral research fellow at Duke University. Now as a Prinicipal Scientist at a biotech company that is orchestrating the power of the epigenome, Brian shares his thoughts on one of the most important questions of his career: "What is the most rewarding aspect of working at a biotech company??" His response below is simple but powerful! 👏🏽👏🏽👏🏽 Watch the full podcast here: https://lnkd.in/eP8ZxGrV

19

4 Comments

Breakthrough in #VirtualScreening! 🚀 University of Washington researchers unveil #RosettaVS, a highly accurate #virtualscreening method, incorporated in #OpenVS, an open-source AI-accelerated platform for drug discovery. These tools outperform existing methods in predicting molecular interactions. 🔬 In tests, OpenVS screened billion-compound libraries against two unrelated targets in under 7 days, achieving impressive hit rates. 🎯 RosettaVS's effectiveness is validated by high-resolution X-ray crystallography, confirming predicted docking poses. This breakthrough could significantly accelerate early-stage drug discovery, paving the way for faster, more efficient pharmaceutical development. Quick Read: https://lnkd.in/dwwQRuyy #Bioinformatics #DrugDiscovery #MachineLearning #ArtificialIntelligence #MedicalResearch #OpenScience #Pharmaceuticals #ScienceNews

17

Breakthrough in #VirtualScreening! 🚀 University of Washington researchers unveil #RosettaVS, a highly accurate #virtualscreening method, incorporated in #OpenVS, an open-source AI-accelerated platform for drug discovery. These tools outperform existing methods in predicting molecular interactions. 🔬 In tests, OpenVS screened billion-compound libraries against two unrelated targets in under 7 days, achieving impressive hit rates. 🎯 RosettaVS's effectiveness is validated by high-resolution X-ray crystallography, confirming predicted docking poses. This breakthrough could significantly accelerate early-stage drug discovery, paving the way for faster, more efficient pharmaceutical development. Quick Read: https://lnkd.in/dwwQRuyy #Bioinformatics #DrugDiscovery #MachineLearning #ArtificialIntelligence #MedicalResearch #OpenScience #Pharmaceuticals #ScienceNews #biotechnology

4

Undermind has dramatically improved the quality and speed of literature search for me and my team. Super excited to see Undermind's progress during YC and equally excited to see the potential impact that their tools can have on the pace of scientific advancement. If you haven't yet tried it, check it out: www.undermind.ai

22

1 Comment

ISoP Community. As the co-chair of this years communications committee, alongside Eleni Karatza, we have some pretty big updates so please read through this post. First and foremost, ACoP 2024 registration is open. Please see the following link (also found below): https://lnkd.in/eNZ2PAcq Note hotel reservations are open and the cut off date for group rate is October 6th. However, as always, if you plan to attend I would suggest booking sooner than later. You can reserve a room by calling 877-800-4888 or through the Reservation site here: https://lnkd.in/ea95yZQS If you haven't checked out the new ISoP website, it looks great! There is a brand new members only section with various benefits. There is a main home feed which contains content/posts from various "communities". Honestly, it feels/looks like an internal Facebook exclusive for our society (I don't mean this in a negative way). You can update your profile, connect with others, and keep up to date with ISoP events/communications. Here is the link the updated website and members section (you must have an active membership): https://lnkd.in/eDVHwgSZ Right now there is only one main community called "Members Connect Community", which contains all ISoP members and acts as a way to make a broad post to everyone. In the future, we could create ISoP SIG specific groups, alumni groups, local ISoP chapters, and others. Even individual working groups within ISoP (e.g., QSP SIG WGs) could create their own channel. Here is the link to the Member Connect Community channel: https://lnkd.in/eBZWrscu If there are any questions feel free to reach out to me directly or comment below! Happy Father's Day!

36

This is a very important read for folks who want to understand the current state of molecular biology and the trend towards RNA tech. This new tech might help with mutations. Now how do we fix gene expression problems? We're working on it. https://lnkd.in/ebu3NkNv Check out the article here: https://lnkd.in/ebu3NkNv

149

2 Comments

"The hype will play out in the long term," James Fraser said. "The hard part of drug discovery (is) it's a pretty long business. You can't take the wet lab part out of it yet. Stuff needs to play out in the clinic." Ron Leuty takes an in-depth look at Profluent and how #AI has become a force in drug development (biologics & small molecule). We are seeing this play out in applications to the QB3 mentoring program. It is striking how many new companies are touting that their use of AI/ML provides a competitive advantage. It's almost more remarkable if a drug discovery company is NOT using AI. And it is very hard to judge the longer-term commercial prospects of a new company whose approach relies on AI/ML. Obviously the quality of the data the models are using is paramount. Anyone with thoughts about how to assess these companies, please weigh in! https://lnkd.in/gaVr3CPj

18

It's been another exciting week for AI in biotech! 🚀 EvolutionaryScale's ESM3 model, simulating 500M years of evolution, is transforming protein design. This joins other recent news on David Baker's Institute for Protein Design raising a whopping $1 billion to further its work. AlphaFold 3 is also pushing the boundaries of protein folding predictions. Still, we must remember that AI is not a magic solution to solve real use cases without having a relevant and efficient feedback loop. True innovation requires mastering both AI tools and hands-on lab skills. A good example is Enzymit which uses tools like ESM and AlphaFold 3 while maintaining a robust experimental feedback loop 💡. This is critical for ensuring success within the short time frames in which startups operate. #AI #Biotech #Innovation

2

Wow - 2 ex-Meta employees launch EvolutionaryScale landing $142m to advance AI in biology. EvolutionaryScale has introduced ESM3, a cutting-edge language model for biology that predicts protein sequences, structures, and functions. This model can simulate millions of years of evolution, generating new proteins with potential applications in medicine, research, and sustainability. ESM3 represents a significant advancement in AI's ability to design and engineer biological systems. There will be comparisons with AlphaFold3, intrigued where this goes... https://lnkd.in/dhtQDtsA #AIinDrugDiscovery #genomics #proteomics

3

1 Comment

Researchers at UCLA & Georgia Institute of Technology unveil #RNAGPT, a multi-modal chat model that simplifies RNA discovery by leveraging extensive literature to streamline complex queries. 🧬 🎯 RNA-GPT integrates RNA sequence encoders with large language models to process user-uploaded sequences. According to the researchers, RNA-GPT effectively addresses complex RNA queries, helping researchers navigate the vast literature and unlock new insights with the power of multi-modal AI. Quick Read: https://lnkd.in/gAeEQJfZ #Bioinformatics #RNA #LLMs #ArtificialIntelligence #gpt #ScienceNews #biotechnology

5

2 Comments

Chai-1 Released by Chai Discovery Team: A Groundbreaking Multi-Modal Foundation Model Set to Transform Drug Discovery and Biological Engineering with Revolutionary Molecular Structure Prediction The Chai Discovery team announced the launch of Chai-1, a groundbreaking multi-modal foundation model designed to predict molecular structures with unprecedented accuracy. This release marks a major advancement in molecular biology and drug discovery, with the model boasting state-of-the-art capabilities across a diverse range of tasks. As a freely available tool, Chai-1 opens new avenues for research and commercial applications, particularly in drug discovery. The core achievement of Chai-1 is its ability to predict complex molecular interactions involving proteins, small molecules, DNA, RNA, and even covalent modifications. This comprehensive scope makes it one of the most versatile tools for molecular structure prediction today. Unlike previous models, which often required multiple sequence alignments (MSAs) for effective predictions, Chai-1 can operate in single-sequence mode without significant loss of accuracy. This breakthrough enables users to predict biomolecular structures more efficiently, particularly when working with multimers. In benchmark tests, Chai-1 demonstrated a 77% success rate on the PoseBusters benchmark, outperforming AlphaFold3, which achieved a 76% success rate. Furthermore, Chai-1 achieved a Cα LDDT (Local Distance Difference Test) score of 0.849 on the CASP15 protein monomer structure prediction set, surpassing the performance of the ESM3-98B model, which scored 0.801. These results place Chai-1 at the cutting edge of molecular structure prediction, challenging the dominance of existing tools like AlphaFold.... Read our full take on this: https://lnkd.in/gzUcbjyU Technical report: https://lnkd.in/gCupqXBu GitHub: https://lnkd.in/gZQqTRaa Try it here: https://lnkd.in/gdm9C9jz Chai Discovery Joshua Meier Vinicius de Freitas Reis Jack Dent

30

1 Comment

This study examines the binding dynamics of E3 ligases (VHL, CRBN, and cIAP) with proteins of interest, focusing on dynamics, cooperativity, selectivity, linker length, and PROTAC conformations. The influence of interface residues and linker lengths on specific conformations for target proteins and E3 ligases is highlighted. Utilizing molecular dynamics and steered molecular dynamics simulations, the study provides comprehensive parameters on the behavior and stability of diverse ternary complexes. These insights are crucial for designing PROTACs targeting disease-causing proteins and advancing the development of degradable ternary complexes for therapeutic applications.

45