NVIDIA Clara™ for Drug Discovery is a collection of GPU-accelerated and optimized frameworks, applications, generative AI platforms, and pretrained models for AI drug discovery pipelines. Built to support cross-disciplinary workflows, Clara for Drug Discovery helps computational biologists, computational chemists, and AI drug discovery researchers and scientists understand disease mechanisms and get drugs to market faster.
Deep learning models based on transformer architectures are poised to accelerate every phase of drug discovery. From large language models (LLMs) that generate de novo proteins, small molecules, and the 3D structure of a protein to molecular dynamic simulations, novel deep learning techniques are changing the way that scientists explore the ever-expanding chemical and biological research space.
Credit: Mahendra awale, CC BY-SA 3.0 https://creativecommons.org/licenses/by-sa/3.0 , via Wikimedia Commons
NVIDIA BioNeMo™ Service is for generative AI drug discovery pipelines, offering domain-specific, state-of-the-art biomolecular models at supercomputing scale through cloud APIs.
Drug discovery spans many workflows, from exploring the chemical universe and predicting protein structures to scanning drug candidates and simulating molecules. Drive breakthroughs in these critical research areas with the powerful tools available in the NVIDIA NGC™ catalog.
Transformer-based LLMs are creating new possibilities for real-time exploration of the chemical and biological space. BioNeMo is a supercomputing platform, built on NeMo™ framework, for training and inferencing biomolecular LLMs and to help scientists quickly identify candidate therapeutics. It contains AI models for predicting protein and small molecule properties (ESM-1, ESM-2, MegaMolBART, MoFlow), protein generation (ProtGPT2), pose prediction (DiffDock), and 3D protein structure prediction (OpenFold, AphaFold2, ESMFold).
The 3D structure of a protein can now be predicted from its primary acid sequence using LLMs such as OpenFold, ESMFold, and AlphaFold2. These models are in BioNeMo Service, enabling exploration of protein structures in seconds.
Deep learning-based approaches like RELION are powering high-throughput automation of cryogenic electron microscopy (cryo-EM) for protein structure determination. RELION implements an empirical Bayesian approach for analysis of cryo-EM to refine singular or multiple 3D reconstructions as well as 2D class averages.
To understand protein structures with atomistic detail, tools like MELD can be used to infer structures from sparse, ambiguous, or noisy data. MELD harnesses data in a physics-based, Bayesian framework for improved protein structure determination.
Image courtesy of Rommie Amaro and the University of California San Diego
With AI and accelerated computing, millions of drug candidates can be screened against a rigid protein target. AutoDock is a growing collection of methods for computational docking and virtual screening for use in structure-based drug discovery and exploration of the basic mechanisms of biomolecular structure.
GPU-powered molecular dynamics frameworks can simulate the fundamental mechanisms of cells and calculate how strongly a candidate drug will bind to its intended protein target. Machine-learned potentials, which show promise for quantum mechanical-level accuracy, energies, and forces, are fundamentally changing molecular simulation.
Clara for Drug Discovery includes a variety of tools and frameworks for molecular simulation, including GROMACS, NAMD, Tinker-HP, VMD, TorchANI, and DeePMD-Kit.
Clara Discovery is optimized to run on NVIDIA DGX™ A100, the world’s most advanced AI system delivering five petaFLOPS of performance. Purpose-built for all accelerated computing workloads at scale, DGX A100 provides researchers the fastest time-to-solution and offers IT a unified, easy-to-deploy infrastructure to support the next generation of drug discovery.
View HPC Application Performance
View Deep Learning Framework Performance
Read the Pharma POD Whitepaper
Learn About Schrodinger’s Advanced Computational Platform
CPU vs GPU: NVIDIA Clara Parabricks, Relion, Autodock-GPU, NVIDIA RAPIDS, Amber, NAMD, VMD, Gromacs, NVIDIA Clara Imaging, BERT Training