During the GTC conference in March 2023, NVIDIA launched BioNeMo Cloud, a series of generative AI cloud services within its AI foundation suite. Building on this innovation, the company now seamlessly integrates the testing version of BioNeMo Cloud API into a platform tailored for drug discovery workflows, as announced today at the J.P. Morgan Healthcare Conference in San Francisco. Initially, BioNeMo introduced tools such as DeepMind's AlphaFold2, MIT's DiffDock, ESMFold and Meta's ESM2, Cornell University's MoFlow, and ProtGPT-2. The updated cloud API will now cover foundational models from three different sources: internal models developed by the company, such as MolMIM for facilitating small molecule generation; open-source models from global research teams optimized by NVIDIA, such as OpenFold for protein prediction AI; and models created by partners, such as Recursion's Phenom-Beta for embedding cell microscopy images. "Healthcare is inherently complex. So our goal is to simplify these models for researchers, allowing them to fine-tune on proprietary data, run AI model inferences through web browsers or cloud APIs, and access pre-trained models for drug development," said Kimberly Powell, Vice President of NVIDIA Healthcare. Powell, who has worked for the tech giant for over 15 years, leads the company's efforts in leveraging GPU computing and deep learning to advance imaging and life sciences. The company also announced that Amgen, a California-based biopharmaceutical multinational, plans to utilize generative AI for drug discovery. Amgen will employ an AI system called Freyja, powered by an NVIDIA DGX SuperPOD, to analyze a vast human dataset at its deCODE Genetics headquarters in Iceland. The system aims to create a human diversity atlas for the discovery of drug targets and disease-specific biomarkers, facilitating diagnostic monitoring. Additionally, Freyja will contribute to the development of AI-driven precision medicine models, leveraging the power of 248 H100 Tensor Core GPUs across 31 NVIDIA DGX H100 nodes for accelerated research. Behind the scenes, the BioNeMo API now provides access to state-of-the-art models, including Phenom-Beta from Recursion, a clinical-stage biotech company supported by NVIDIA. This AI model is designed as a visual transformer specifically for extracting biologically meaningful features from images captured through cell microscopy. The primary focus is on utilizing AI to identify and understand key features within cellular structures, helping researchers gain valuable insights into cell functionality and responses to stimuli such as drug candidates or genetic engineering. Phenom-Beta excels in image reconstruction tasks, a critical indicator of model proficiency. The model was trained on Recursion's RxRx3 dataset using the BioHive-1 supercomputer, based on the NVIDIA DGX SuperPOD reference architecture. To enhance model development, Recursion is expanding its supercomputer by adding over 500 NVIDIA H100 Tensor Core GPUs, aiming to create one of the most powerful supercomputers owned by a pharmaceutical company. On the other hand, NVIDIA's proprietary model, MolMIM, generates small molecules while allowing users to have better control over the AI generation process by identifying new molecules with specified characteristics and constraints similar to a given reference molecule. MolMIM is trained using a method called Mutual Information Machine (MIM) learning and creates fixed-size representations of different types of molecules. In addition to Amgen, several other companies are utilizing NVIDIA BioNeMo for biological, chemical, and genomic research. For example, Terray Therapeutics integrates BioNeMo Cloud API into its multi-target structure binding model development. Innophore and Insilico Medicine apply BioNeMo to computational drug discovery, with Innophore integrating it into the Catalophore platform and Insilico using BioNeMo in the early stages of drug discovery within their generative AI pipeline. Furthermore, OneAngstrom and Deloitte are leveraging BioNeMo Cloud API to build AI solutions. OneAngstrom focuses on molecular design on the SAMSON platform, while Deloitte integrates BioNeMo into the Quartz Atlas AI platform, which utilizes NVIDIA DGX Cloud for scientific research. This integration enhances data connectivity and generative AI capabilities, propelling pharmaceutical researchers into a new era of accelerated drug discovery. Recently, in November of last year, Genentech, a subsidiary of the Roche Group, and NVIDIA signed a multi-year research collaboration to strengthen the company's machine learning algorithms specifically for AI applications in drug discovery, utilizing NVIDIA's DGX Cloud platform. In the realm of healthcare initiatives, NVIDIA is making a real-life impact, and other tech giants are not far behind. Isomorphic Labs, the drug discovery branch of Alphabet, announced today that it is collaborating with pharmaceutical giants Eli Lilly and Novartis to develop small molecule therapeutic approaches for multiple targets. Just two months ago, Isomorphic and Google DeepMind released an updated version of AlphaFold 2, which can now predict the structure of almost all molecules in the Protein Data Bank (PDB), a comprehensive database of three-dimensional biological molecules. It has expanded its capabilities to include small molecules, proteins, nucleic acids, and molecules with post-translational modifications. AlphaFold has already found applications in various real-life scenarios, including malaria research, liver cancer, COVID-19 vaccine development, and gene therapy. On the other hand, OpenAI, supported by Microsoft, is collaborating with WHOOP to develop a GPT-4-driven personalized health coach. They are also partnering with Summer Health to generate medical records from detailed observations using GPT-4. Similarly, Apple and Oracle are investing in AI implementation in healthcare. Surprisingly, Meta has maintained a noticeable silence in this critical field after disbanding its protein folding team.