Biotechnology

AI-Designed Proteins Let Scientists See Inside Living Cells Like Never Before

A collaboration between Nobel laureate David Baker's lab and researchers at Janelia and EMBL has produced NovoTags — entirely synthetic proteins that bind to brilliant fluorescent dyes, opening a window into the molecular machinery of life with unprecedented clarity.

NovoTags are de novo protein tags designed by artificial intelligence, not evolved in nature. They bind with nanomolar affinity to ultra-bright Janelia Fluor dyes, combining the genetic targetability of fluorescent proteins with the brightness of synthetic dyes.
The system includes far-red, orange, and green variants that can be used simultaneously, allowing researchers to label and track up to 30 different proteins in a single living cell — a massive leap over the 4-5 color limit of conventional methods.
Published in Science, the breakthrough was made possible by AI-driven protein design pioneered by David Baker, who won the Nobel Prize in Chemistry in 2024 for this very approach.

For decades, biologists faced a trade-off: fluorescent proteins were genetically encodable but dim, while synthetic dyes were bright but difficult to attach to specific proteins. NovoTags shatter this compromise. The AI-designed proteins are small, stable, and bind with exquisite specificity to cell-permeable dyes that are among the brightest known. The result is a labeling system that is both precise and brilliant.

The team demonstrated NovoTags across multiple imaging modalities, including confocal microscopy, structured illumination microscopy, and fluorescence lifetime imaging. In each case, the tags outperformed existing fluorescent proteins in brightness and photostability while maintaining the ability to target specific proteins through genetic fusion. The far-red variant is particularly valuable because it allows imaging deeper into tissue with less background autofluorescence.

Beyond its immediate utility for cell biology, NovoTags represents a broader validation of AI-driven protein design. If algorithms can produce proteins with functions that evolution never discovered, the approach could be applied to drug delivery, biosensors, and synthetic biology. For now, the tags are already being shared with the research community, and labs around the world are expected to begin using them within weeks.