CZI and NVIDIA collaborate to accelerate life science research through virtual cell models

Today, the Chan Zuckerberg Initiative (CZI) and NVIDIA announced an expanded collaboration to accelerate life science research by driving development and adoption of virtual cell models through tools, data, models, and benchmarks delivered through CZI's virtual cells platform (VCP). Core to this collaboration is an effort to scale biological data processing to petabytes of data spanning billions of cellular observations, enabling next-generation model development that will unlock new insights about human biology. 

The burgeoning field of virtual cell model development is rapidly evolving with the continued generation of large-scale, multi-modal biological datasets that are ripe for AI-driven insights about health and disease. CZI's VCP lowers the barriers for biologists to apply AI to specific biological tasks while enabling AI/machine learning researchers to rapidly iterate and improve model quality. AI and life science leaders like CZI, combined with NVIDIA's AI and accelerated computing expertise, can supercharge the development of virtual cell models. This includes scaling harmonized data within the VCP, providing the infrastructure and technical capacity to optimize training, and further expanding the scope and accessibility of datasets and models available to the scientific community. 

Our collaboration with NVIDIA is a significant step forward in our mission to harness the power of AI for biological discovery. By combining our expertise in biological data generation and model development with NVIDIA's leadership in accelerated computing, we can provide researchers with the infrastructure and tools they need to make novel discoveries about human biology and ultimately solve disease."

Ram Balasubramanian, VP of science technology at CZI

The new tools and resources being announced as part of the expanded collaboration include:

• Scaling Data Processing: Generation and harmonization of large biological datasets have proven to be a bottleneck for AI applications for science. To increase the scale and breadth of data available to the scientific community, CZI and NVIDIA will collaborate to scale biological data processing to billions of data points, leveraging GPU-accelerated tools. This will enable CZI to produce large-scale datasets for virtual cell model development and create robust data platforms to support large-scale data publishing, exploration, and ecosystem development.

• Accelerating Cutting-Edge Model Development: Multi-modal, multi-scale, and multi-domain modeling will be critical to achieve comprehensive and holistic insights that reflect the complex, interconnected nature of a physical cell. CZI's virtual cell modeling teams have developed state-of-the-art virtual cell models such as rBio, GREmLN, and TranscriptFormer. As CZI scales data generation, they will also increase the scale and scope of their virtual cell models across multiple scales of biology. CZI and NVIDIA will work together to scale and accelerate virtual cell model development using NVIDIA technology.

• Accessing AI Models and Data through CZI's virtual cells platform (VCP): CZI's VCP is an open-source platform that makes it easy to find and access data, models, and AI-powered biological analysis tools built by CZI and the broader scientific community to accelerate biological discovery. CZI's VCP hosts state-of-the-art virtual cell models developed by CZI and collaborators, including NVIDIA Clara Open Models such as MONAI-based imaging models and CodonFM, an RNA foundation model, which are being offered on the platform for the first time, creating a unified ecosystem for open, reproducible biological AI research. The platform also includes cz-benchmarks, an open source Python package co-developed with NVIDIA that allows model developers to spend less time figuring out how to evaluate their models and more time improving them to solve real biological problems.

"We are collaborating with CZI to provide advanced AI computing infrastructure, domain-specific software, and deep expertise in data processing, model scaling and deployment - all to enable a new generation of AI-powered models that accelerate discovery in biology," said Rory Kelleher, senior director and global head of business development for life sciences at NVIDIA. "By supporting CZI's efforts to create open, community-driven resources, we are helping to ensure that these powerful tools can be used by scientists everywhere to solve some of the most complex challenges in medicine."

Researchers can access rich datasets, cutting-edge AI models, tools, and benchmarks today through CZI's open and freely available virtual cells platform.