Rice Team Generates 10M Protein Data Points in 3 Days

Scientists just cracked a major roadblock that was holding back the next generation of life-saving proteins.

Researchers at Rice University developed a breakthrough system called Sequence Display that generates over 10 million data points about proteins in a single experiment. The entire process takes just three days, solving what had been one of artificial intelligence's biggest challenges in designing better proteins for medicine and research.

The problem was simple but massive. Proteins are made of amino acids, and even a small protein with just 50 amino acids has more possible combinations than scientists could ever test in a lab. That's a number with 65 zeros after it, five times more zeros than a trillion.

AI seemed like the perfect solution for sorting through these possibilities. But the technology had a catch: it needed huge amounts of quality data to learn from, and that data simply didn't exist for optimizing what proteins actually do.

"One of the biggest bottlenecks in AI-guided protein engineering is not coming up with machine-learning models," said Han Xiao, Rice University professor and director of the SynthX Center. "It is generating the right and enough experimental data to train them."

The team's solution was elegantly simple. They created thousands of variations of a protein and attached a blank DNA barcode to each one. As each protein variant performed its job, a special editor changed the barcode based on how well it worked. The most active proteins got the biggest barcode changes.

Next-generation sequencing then read all those barcodes like scanning items at a grocery store. The result: millions of data points showing exactly which protein versions worked best.

Graduate student Linqi Cheng, who led the study published in Nature Biotechnology, tested the system on a small CRISPR-Cas protein. The protein was useful because of its size, but it could only cut limited stretches of DNA. After running through Sequence Display, the AI model successfully predicted mutations that dramatically improved the protein's ability to target a wider variety of DNA.

The team repeated their success with several other proteins, proving the approach works across different types.

The Ripple Effect

This breakthrough changes the relationship between AI and lab work. Instead of AI replacing experiments, it builds on them. Scientists can now generate the training data AI needs to search through possibilities that would take lifetimes to test by hand.

The implications stretch far beyond the lab. Better protein engineering means more effective medicines, improved research tools, and faster development of treatments for diseases. Each protein that gets optimized through this process could become a new therapy or diagnostic tool helping patients around the world.

The system creates a practical framework that any protein engineering lab can use. It doesn't require rare equipment or impossible expertise, just a smart combination of existing tools used in a new way.

What once took years of guesswork and limited results now takes three days and produces millions of answers.