One billion years of protein evolution reveals surprising design flexibility

One billion years of protein evolution reveals surprising design flexibility
by Hugo Ritmico
Madrid, Spain (SPX) Jul 25, 2025
Proteins, the molecular machines of life, are constructed from 20 amino acids in seemingly infinite combinations-up to 10^78 possible sequences for a 60-residue protein. Yet only a tiny subset fold into stable, functional shapes. A new study published in Science reveals that the rules guiding these folds may be far more forgiving than previously assumed.

Scientists from the Centre for Genomic Regulation (CRG) in Barcelona and the Wellcome Sanger Institute in the UK studied the FYN-SH3 protein domain, a component found in many human proteins. By generating and testing hundreds of thousands of FYN-SH3 variants, they discovered that most combinations still folded properly and retained function. Contrary to long-standing assumptions, only a few amino acids in the core were truly essential to maintain structure.

"Our data challenges the dogma of proteins being a delicate house of cards. The physical rules governing their stability is more like Lego than Jenga, where a change to one brick threatening to bring the entire structure down is a rare, and crucially, predictable phenomenon," said Dr. Albert Escobedo, postdoctoral researcher at CRG and lead author of the study.

Using machine learning, the team trained a model on their vast dataset to predict which SH3 sequences would remain stable. When tested against over 51,000 naturally occurring SH3 domains across species from bacteria to humans, the algorithm correctly identified nearly all as stable-even when sequences were less than 25 percent identical to the human version.

"Evolution didn't have to sift through an entire universe of sequences. Instead, the biochemical laws of folding create a vast, forgiving landscape for natural selection," added Dr. Escobedo.

This breakthrough has wide implications for protein engineering. Current methods rely on making incremental changes and screening each variant-a slow and costly process. The new findings suggest that larger, riskier design changes are viable and predictable, significantly reducing trial-and-error phases in developing enzymes, vaccines, or drug candidates.

For instance, redesigning protein surfaces to avoid immune reactions has historically been painstaking due to concerns about destabilizing the protein core. With this new predictive model, scientists can make numerous simultaneous changes in silico and move directly to testing promising candidates.

"The ability to predict and model protein evolution opens the door to designing biology at industrial speed, challenging the conservative pacing of protein engineering," said ICREA Professor Ben Lehner, senior author of the study and researcher at both CRG and the Wellcome Sanger Institute.

Research Report:Genetics, energetics, and allostery in proteins with randomized cores and surfaces