← Back to Home

Taking AI Design Beyond the Computer: U.S. Invests in an Open-Access Cloud Laboratory for Protein Engineering

Northwestern University has received $20 million in funding over four years from the U.S. National Science Foundation to connect AI design, cell-free protein synthesis, and robotic testing into a remote experimental workflow. Although the goal is to build a large open database, the facility has yet to be constructed, and its performance and usage rules remain unclear.

By SURL BioNews

Generative AI for proteins can propose thousands of designs on a computer, but the real bottleneck lies at the lab bench: whether candidate proteins can be produced, whether they are stable, and whether they can perform their intended functions must still be validated one by one. Northwestern University in the United States has now received $20 million to transform this costly and slow process into an automated experimental service that researchers can use remotely.

The funding will be provided over four years by the U.S. National Science Foundation (NSF) to establish the AI-Driven, Rapid, Experimental Automation Machine Cloud Lab (DREAM Cloud Lab). The facility will be housed within the Synthetic Biology Foundry under Northwestern University’s Center for Synthetic Biology and led by Julius Lucks. Team members come from Northwestern University and Stanford University, spanning synthetic biology, protein engineering, computer science, and statistics.

The planned workflow begins with a cloud interface: users describe the function they want a protein to perform, AI models propose candidate designs, and, after biosafety and biosecurity reviews, robotic systems construct and test them. The team plans to use cell-free protein synthesis technology, directly using molecular machinery outside cells to produce proteins. Experimental results will then be fed back into the models, forming an iterative “design, build, test, learn” cycle.

The biomedical value of this system lies not only in rapidly generating candidate sequences, but also in providing the standardized experimental feedback most lacking in AI protein engineering. If researchers can compare protein expression levels, stability, and function on the same platform, the data could be used to improve design models for therapeutic proteins and other biotechnologies. The university estimates that within four years it will synthesize and characterize more than 300,000 proteins, generate up to 30 million data points, and openly release the data and AI models.

DREAM is also one of the NSF’s first 20 “programmable cloud laboratories.” The NSF announced an investment of $380 million to establish a national network spanning biology, chemistry, materials, and electronics, while the Astera Institute may provide up to an additional $20 million in support. The shared aim is to enable researchers to remotely orchestrate experiments and produce reusable data suitable for AI training. Northwestern University’s node will focus on protein engineering and will establish an ethics and responsible innovation committee to participate in governance.

However, what has been announced so far consists of construction goals rather than completed technical validation. The university has not yet provided model hit rates, robotic experimental throughput, data quality standards, or queuing and fee mechanisms for external users. The specific scope of “open access,” as well as how protein designs with potential dual-use applications will be reviewed, also remains to be formalized. Whether the laboratory can truly shorten drug discovery cycles will ultimately depend on whether the data it generates are reliable and reproducible, and whether they can be translated into proteins with practical functions in independent research.

References

  1. Northwestern University
  2. Northwestern Now
  3. U.S. National Science Foundation