ArcInstitute/arc-virtual-cell-atlas

What it solves

It provides a massive, curated collection of high-quality open datasets to accelerate the development of virtual cell models, removing the barrier of fragmented or low-quality biological data.

How it works

The project hosts a vast library of observational and perturbational cellular data—covering over 602 million cells—on the Google Cloud Marketplace. It integrates major datasets like Tahoe-100M and the AI agent-curated scBaseCount, allowing researchers to download specific subsets of data (partitioned by species, release, and quantification) using Google Cloud tools.

Who it’s for

Researchers and AI developers working on biological modeling, single-cell analysis, and the creation of virtual cell simulations.

Highlights

• Massive Scale: Contains data from over 602 million cells.
• Curated Quality: Includes datasets specifically assembled and curated (some via AI agents) for model training.
• Cloud-Native Access: Hosted on Google Cloud Marketplace with a free tier for subscribed projects (up to 2TB of egress per month).