NSF Funding for Caltech-Led "Cloud Laboratory" Leverages AI to Explore Chemical Dark Matter: A Conversation with Hosea Nelson
The National Science Foundation awarded $17 million to Caltech-led ELECTRA, an AI-driven cloud laboratory, to map millions of unknown molecular structures known as 'chemical dark matter' using automated experiments and electron diffraction.
Caltech chemist Hosea Nelson will lead a $17 million NSF-funded project to build ELECTRA, a cloud laboratory that combines robotics and AI to remotely analyze molecular structures. The initiative targets 'chemical dark matter'—millions of unknown molecules in nature—using microcrystal electron diffraction, a technique Nelson’s lab adapted from structural biology. The platform aims to accelerate discovery across chemistry, biomedicine, and materials science by enabling rapid, large-scale structural elucidation.
ELECTRA will focus on biomedical applications, including natural products like taxol and aspirin derivatives, which remain understudied despite their potential for drug development. The project also addresses gaps in basic biology and geoscience, where unknown molecular structures hinder progress in fields such as microbiome research and rare earth mineral analysis. Collaborators span multiple disciplines, leveraging expertise in synthesis, structural biology, and polymer science to build a versatile platform.
The cloud laboratory will automate experiments using high-throughput robots, reducing the time required to determine molecular structures from weeks to under a minute. AI, including contributions from Caltech’s Katie Bouman and Yisong Yue, will process and interpret the vast datasets generated. The system will operate as an open-access resource, allowing researchers nationwide to submit samples and access shared data via a cloud-based interface.
A core challenge is the lack of sufficient training data for AI models in chemistry, which requires generating massive datasets through automated experimentation. ELECTRA will crowdsource samples from across the U.S., creating a centralized repository of molecular structures. The project aligns with the U.S. government’s Genesis Mission to develop AI-driven autonomous laboratories, positioning ELECTRA as a transformative tool for scientific discovery.