NVIDIA’s Post

The future of protein design is here with EvolutionaryScale's groundbreaking ESM3 model. This 98 billion parameter AI system simulates millions of years of evolution to create novel proteins, accelerating scientific breakthroughs in fields from drug discovery to environmental sustainability[1][2][3]. In a remarkable demonstration, ESM3 generated a new Green Fluorescent Protein (GFP) variant in a fraction of the time it would take natural evolution, exhibiting a 58% dissimilarity to the nearest known fluorescent protein[4][5]. ESM3's potential applications are vast - from designing new therapeutics and antibodies, to engineering proteins for carbon capture and breaking down plastics. Its interactive nature allows researchers to prompt it for specific protein designs, acting as a virtual collaborator in solving complex biological challenges[1][2][3]. To make ESM3 accessible to researchers worldwide, EvolutionaryScale has partnered with AWS to make the full model available, and secured over $142 million in seed funding from investors including Lux Capital, Amazon, NVIDIA, and more[1][5]. The future is now. Discover the power of ESM3 and how it will revolutionize protein design and scientific research. Sources: [1] NVIDIA's Post - LinkedIn https://lnkd.in/dxCCYvq2 [2] ESM3 is a new LLM for new proteins - LinkedIn https://lnkd.in/dmBarNpF [3] Dror Hilman - ESM3's Leap in Protein Modeling Reading… - LinkedIn https://lnkd.in/dHAi6TEX [4] ESM3: A Powerful Protein Language Model Simulating 500 Million Years ... https://lnkd.in/dpnkXQi4 [5] Aniket Deshpande on LinkedIn: EvolutionaryScale's ESM3 model is ... https://lnkd.in/dGWU2BrT #AI #ProteinDesign #DrugDiscovery #Sustainability #GreenFluorescentProtein #FutureofScience #TechInnovation #EvolutionaryScale #ESM3

Bioinformatics powered by ML will be an interesting trend to watch out for in the coming few years

The work being done and recent breakthroughs on protein design is revolutionary and can certainly lead up to humanity searching for the answers of their origin and the answer to the question of "Are we alone". Processing power and neural network and Fnet technology is most certainly the future

Learn more about ESM3, which used NVIDIA H100 GPUs for the most compute ever put into a biological foundation model. The 98 billion parameter EMS3 model uses roughly 25x more flops and 60x more data than its predecessor, ESM2. #AI

Super interesting seeing this from NVIDIA. 😎 I personally love seeing Generative AI concepts being used to do things like this and accelerating Bio research and healthcare. #generativeai #ai #medical https://lnkd.in/dUET9b2P

🌐💻 Bridging the Divide: Neuromorphic Computing Integrates Digital and Biological Realms 🧠💻 Neuromorphic Computing: Mimics human brain processing 🧠, promising a revolution in AI 🤖 and robotics, originated by Carver Mead in the 1980s. 🖥️💻 Hardware & Software: Involves neuromorphic chips like Intel's Loihi 💻 and algorithms modeling brain functions, 🧠 enhancing machine learning efficiency and adaptability. ⚡🔋 Impact & Challenges: Offers efficient data processing with lower power consumption 🔋, real-time learning, and biologically inspired learning, but faces scalability and integration challenges. 🔮🤖 Future Prospects: Expected to improve intelligent technologies 🤖, leading to smarter systems and new human-machine interactions 🤝, with ongoing research to overcome current limitations. 🌍🤝 #neuromorphic #computing #AI #technology #innovation #machinelearning #futuretech #digitaldivide #bioinspired

AI docking for real-world virtual screening has arrived London-based startup RECEPTOR.AI announced a new version of its AI docking software, ArtiDock v2.0. The company also published a preprint disclosing details about the model architecture, training, and tuning. Receptor.AI notes that the first generation of AI docking techniques used simple and lightweight model architectures and demonstrated results that were inferior to conventional docking. This issue was mitigated, to some extent, in the second generation of AI docking models by using much heavier architectures. The boost of accuracy, however, came at the expense of very complex architectures, large model sizes, and, as a result, very slow training and inference. ArtiDock exploits the opposite approach by providing a deliberately lightweight and fast model architecture, which is trained on a larger amount of augmented data. This results in better prediction accuracy without compromising the inference speed. ArtiDock uses two sources of augmented data: the PocketCFDM algorithm for generating artificial protein-ligand complexes, which mimic real protein binding pockets in terms of statistical distributions of the non-bond interactions, and the ensembles of representative protein conformations obtained from the impressively massive MD simulations of ~17,000 protein-ligand complexes. ArtiDock 2.0 excels on the PoseBusters v3 dataset, which is deliberately designed to challenge AI docking technologies. The model systematically outperforms all other AI docking techniques, and it also leaves behind conventional docking programs such as Glide, Gold and Vina. The quality metrics of the generated binding poses are unprecedented for AI techniques and go on par with conventional docking, according to comments from the Receptor.AI team. All this is several orders of magnitude faster than the industry standard rivals, which allows using ArtiDock in ultra-high-throughput virtual screening scenarios involving multiple protein conformations and multiple explicit off-targets. In some of the case studies communicated by Receptor.AI, usage of ArtiDock allowed for achieving hit rates of a whopping 40% and discovery of the lead-like compounds with in vivo activity from the single iteration of virtual screening. (link in the comments) ArtiDock is currently being integrated into the NVIDIA hashtag #BioNeMo cloud platform for drug development, which will make it available to a wide range of interested biotech and pharma companies in the near future. #MedicinalChemistry #DrugDiscovery #artificialintelligence #Artidock Video credit: RECEPTOR.AI

📣Breaking News! Neurxcore Introduces Innovative NPU Product Line for AI Inference Applications, Powered by NVIDIA Deep Learning Accelerator Technology   Grenoble, France, October 19, 2023 – Neurxcore, a leading provider of cutting-edge Artificial Intelligence (AI) solutions, today announced the launch of its Neural Processor Unit (NPU) product line for AI inference applications. It is built on an enhanced and extended version of the open-source NVIDIA's Deep Learning Accelerator (Open NVDLA) technology, combined with patented in-house architectures. The SNVDLA IP series from Neurxcore sets a new standard for energy efficiency, performance, and capability, with a primary focus on image processing, including classification and object detection. SNVDLA also offers versatility for generative AI applications and has already been silicon-proven, operating on a 22nm TSMC platform, and showcased on a demonstration board running a variety of applications. #ai #npu #nvdla #inference #artificialintelligence Read the full press release:

Generative AI: Transforming scientific discovery Generative AI has the potential to revolutionize scientific research by enabling new discoveries and accelerating existing ones. This was the key takeaway from a fascinating conversation with Anima Anandkumar, Director of the Machine Learning Research Lab at NVIDIA. She discussed a range of applications for generative AI in science, including: 1. Predicting dangerous new variants of viruses 2. Improving extreme weather forecasting with better risk assessments 3. Accelerating materials design and drug discovery through simulations She highlighted the importance of incorporating scientific knowledge and constraints into AI models for more robust extrapolation. This inductive bias helps generate physically valid predictions. She also emphasized the need for transparency, explainability, and fairness in deploying AI models, especially for sensitive applications like healthcare. #generativeAI #scientificdiscovery #artificialintelligence #machinelearning #datascience

AI docking for real-world virtual screening has arrived London-based startup RECEPTOR.AI announced a new version of its AI docking software, ArtiDock v2.0. The company also published a preprint disclosing details about the model architecture, training, and tuning (link in the comments). Receptor.AI notes that the first generation of AI docking techniques used simple and lightweight model architectures and demonstrated results that were inferior to conventional docking. This issue was mitigated, to some extent, in the second generation of AI docking models by using much heavier architectures. The boost of accuracy, however, came at the expense of very complex architectures, large model sizes, and, as a result, very slow training and inference. ArtiDock exploits the opposite approach by providing a deliberately lightweight and fast model architecture, which is trained on a larger amount of augmented data. This results in better prediction accuracy without compromising the inference speed. ArtiDock uses two sources of augmented data: the PocketCFDM algorithm for generating artificial protein-ligand complexes, which mimic real protein binding pockets in terms of statistical distributions of the non-bond interactions, and the ensembles of representative protein conformations obtained from the impressively massive MD simulations of ~17,000 protein-ligand complexes. ArtiDock 2.0 excels on the PoseBusters v3 dataset, which is deliberately designed to challenge AI docking technologies. The model systematically outperforms all other AI docking techniques, and it also leaves behind conventional docking programs such as Glide, Gold and Vina. The quality metrics of the generated binding poses are unprecedented for AI techniques and go on par with conventional docking, according to comments from the Receptor.AI team. All this is several orders of magnitude faster than the industry standard rivals, which allows using ArtiDock in ultra-high-throughput virtual screening scenarios involving multiple protein conformations and multiple explicit off-targets. In some of the case studies communicated by Receptor.AI, usage of ArtiDock allowed for achieving hit rates of a whopping 40% and discovery of the lead-like compounds with in vivo activity from the single iteration of virtual screening. (link in the comments) ArtiDock is currently being integrated into the NVIDIA #BioNeMo cloud platform for drug development, which will make it available to a wide range of interested biotech and pharma companies in the near future. #MedicinalChemistry #DrugDiscovery #artificialintelligence #Artidock Video credit: RECEPTOR.AI