Claude Science and Nvidia BioNeMo Cut Genomic Analysis From Hours to Seconds

Anthropic and Nvidia have combined their technologies in a way meant to meaningfully cut the time it takes to discover new drugs. The Claude Science research platform now has native integration with the Nvidia BioNeMo Agent Toolkit, letting scientists request complex biomolecular analyses in plain natural language while the system automatically runs the right models and libraries across distributed GPU resources.

Claude Science entered public beta at the end of June as a digital workbench for scientists, described by Anthropic as a place where a researcher runs an entire research process through conversation with an AI agent. The new BioNeMo integration adds Nvidia's computing infrastructure specialized for life sciences, from genomics to molecule design.

How the Process Works

In a typical scenario, a researcher starts by identifying a specific cancer-causing antigen mutation and asks Claude to design potential inhibitors that block the mutated protein's activity. Using the BioNeMo Agent Toolkit and Nvidia NIM microservices, the system runs large-scale prediction, optimization and validation of drug candidates, returning results in a timeframe that previously would have required days of a team's work.

That speed is the core of the announcement. Nvidia notes that its Parabricks genomic analysis tool, which previously required hours of computation per sample, now returns results in a few minutes. An even more striking jump comes from RAPIDS-singlecell, a library for analyzing single-cell data, where processing a 1.3 million-cell dataset was cut from 52 minutes to 25 seconds.

The Speedup in Numbers

The most striking result comes from cheminformatics. The nvMolKit tool, used for chemical similarity searches and molecule conformer generation, speeds up some operations by as much as three thousand times compared to traditional computational methods. For a lab designing hundreds of variants of a potential drug, that means being able to test in a single day what previously took weeks.

Claude Science integrates with the Nvidia BioNeMo Agent Toolkit as a resource that scientists can access directly within their workflow - from the NVIDIA blog

Who Is Already Using It

BioNeMo's adoption in the pharmaceutical industry is already impressive, even before the Claude Science integration. Nvidia says 18 of the world's 20 largest pharmaceutical companies use the toolkit in production environments, meaning the new conversational analysis capabilities are landing immediately with organizations running real drug research, not just academic labs testing novelties.

The integration also includes access to specific biomolecular models. Evo 2 handles analysis and generation of genetic sequences, Boltz-2 predicts protein structure and interactions, and OpenFold3 predicts biomolecular structures. All three are available directly within Claude Science as tools the agent can call on its own in response to a researcher's question posed in plain language.

What This Means for Research and Poland

For Polish research institutes and biotech companies that lack GPU infrastructure on the scale of major pharmaceutical corporations, this kind of integration lowers the barrier to advanced biomolecular analysis. Instead of building their own computing pipelines for genome analysis or molecule design, a research team can use a ready-made toolset available through the cloud, paying for compute time rather than investing in hardware.

The pace at which Anthropic is expanding Claude Science suggests the company views science as one of the priority use cases for its models, alongside coding and office work. Just weeks after its beta launch, the platform has already landed its second major technology partnership, this time with the company that controls most of the market for GPUs used in scientific computing.

Sources: NVIDIA Blog (blogs.nvidia.com), Artificial Intelligence News (artificialintelligence-news.com), iMagazine (imagazine.pl)