Drug discoverers are among customers Nvidia has pinpointed for its recently launched catalog of pretrained, customizable workflows enabling enterprise-level users to make their own artificial intelligence (AI) applications.
NVIDIA NIM™ Agent Blueprints are reference workflows designed to help drug discoverers and other customers build and deploy generative AI applications for uses that include virtual screening, information retrieval, and even customer service avatars.
The Blueprint incorporates NVIDIA NIMs or optimized cloud-native “microservices” designed to let developers accelerate deployment of generative AI models anywhere—whether through local workstations, on-premises data centers, cloud services, or GPU-accelerated workstations. (NIM stands for “NVIDIA Inference Microservices”).
By enabling biopharmas to shift from traditional fixed database screening to generative AI-driven molecule design and pre-optimization, NIM Agent Blueprints are designed to help researchers design better molecules faster. That, according to Nvidia, represents a paradigm shift in the drug discovery process—especially in the transition of “hit” compounds into “lead” compounds optimized for further development.
“A lot of the drug companies I visit have 60 million molecules in their library. And that is what they’re screening against, that static 60 million molecules. But there are 1060 potential molecules in the chemical space that might be a therapy,” Kimberly Powell, Nvidia’s vice president of healthcare, told GEN Edge.
“The paradigm shift here is what generative AI is doing, and especially what the MolMIM system is doing, is using the generative effect, intelligently searching through chemical space so that a molecule that maybe the world has never synthesized before can have the features that you really care about,” Powell said.
For drug developers, Nvidia says, a NIM Agent Blueprint called generative virtual screening helps them deliver on AI’s longtime promise of reducing the time and cost of developing new therapies, by accelerating the virtual screening of small molecules using generative models.
Improving “hits” via three AI models
The Blueprint identifies and improves virtual “hit” compounds—identified through screening as having potential biological activity—in a smarter and more efficient way. At the core of generative virtual screening are three essential AI models:
- AlphaFold2, an AI model for protein folding developed by Google DeepMind. AlphaFold2 can predict 3D structures of proteins from amino acid sequences with atomic-level accuracy.
- DiffDock, the molecular docking model designed to predict the binding structure of a small molecule ligand to a protein, while simultaneously optimizing for multiple properties, such as high solubility and low toxicity.
- MolMIM, a generative chemistry model that generates drug candidates optimized for properties defined by users. MolMIM can also design molecules that are optimized to bind to a specific protein target.
Each AI model is packaged within NIMs that integrate the microservices into a flexible, scalable, generative AI workflow. The Blueprint pre-optimizes molecules for desired therapeutic properties, using a generative AI approach.
“Virtual screening is still just one piece of drug discovery. But we are working on models that span all the way from target discovery to lead identification. And we’re going to be building out Blueprints along that drug discovery process,” Powell said.
“Computer aided drug discovery is really going to see generative AI injected into it all along the way,” she explained. “A lot of times, computer aided drug discovery has been thought of in the lead identification optimization, where we’re doing a lot of simulation. But now we’re using a lot of computational methods all the way from target ID to the lead optimization.”
Other NVIDIA NIMs focused on drug discovery include:
- ESMFold, a “Transformer” model—a neural network that learns context and thus meaning by tracking relationships in sequential data—which can accurately predict protein structure based on a single amino acid sequence.
- Parabricks DeepVariant (the tool behind the Universal Variant Calling Microservice), a deep learning model designed to help identify variants in short- and long-read sequencing datasets. Parabricks is designed to deliver 50x speed improvement for variant calling in genomic analysis workflows compared to the original or “vanilla” DeepVariant implementation designed to run on central processing units or CPUs.
“We’re up to four or five generally available NIMs, and we have a whole bunch of other drug discovery and healthcare NIMs that are in preview. We’re going to be very active in the announcement and deployment of these applications. Every month, you’re going to see a new rich offering of NIMs and Blueprints,” Powell said.
Digital human, PDF data extraction
In addition to the drug discovery Blueprint, other NIM Agent Blueprints include a digital human workflow for uses ranging from digital health to customer service; and a multimodal PDF data extraction workflow for enterprise retrieval-augmented generation (RAG) designed to generate more accurate responses from vast quantities of business data.
According to Nvidia, RAG can read images from any PDF, and furnish insights based on what it sees.
“As the healthcare industry through biomedical research, everything we do with insurance companies, all the patient and doctor interactions, there’s PDFs everywhere that have a massive amount of useful information into it. Now we’re going to be able to extract it out and summarize it,” Powell said.
The digital human workflow, which can also be applied in digital health, uses customized avatars capable of automatic speech recognition—such as Nvidia’s interactive digital human, named “James.” Human speech is converted into text that goes into a language model that goes from there into the RAG system, returns into speech synthesis, and activates a user’s avatar.
“If you do that complete loop, you now have the representation of truly a digital human who can understand, reason, and respond, and also use the Audio2Face NIM,” Powell said. “You can meet James, and when a text to speech comes back, you’re actually showing a different emotion on the face, and you’re able to have a much more engaging conversation.”
The Blueprints are free for developers to download and can be deployed with the NVIDIA AI Enterprise software platform.
Biopharmas—including all of the top 20 in the space, according to Nvidia—are accessing NVIDIA NIM Agent Blueprints through global system integrators and technology solutions providers that include Accenture, Deloitte, SoftServe and World Wide Technology (WWT) are bringing to enterprises worldwide. Cisco, Dell Technologies, Hewlett Packard Enterprise and Lenovo are offering full-stack NVIDIA-accelerated infrastructure and solutions to speed NIM Agent Blueprints deployments.
Accenture plans to customize a NIM Agent Blueprint to the specific needs of drug development programs by partnering with biopharmas to optimize the molecule generation step within the MolMIM NIM.
Amazon Web Services’ AWS HealthOmics—a service designed to help biopharmas and healthcare systems store, query, and analyze genomic, transcriptomic, and other omics data—is making available all three NIMs that make up the Blueprint, with the goal of streamlining the integration of AI into existing drug discovery workflows, Nvidia said.
