Particle.news
Download on the App Store

PNAS Study Finds Ancestry Strongly Limits Protein Evolution

The authors warn that AI protein design trained on known sequences may not reach the far larger unknown space without new experimental data.

Overview

  • The peer-reviewed study in PNAS by teams at OIST, ISTA, the University of Vienna, and CAB examines how proteins spread through the vast space of possible amino acid sequences.
  • The researchers find ancestry is the strongest brake on diversification, leaving many protein families tightly clustered and low in effective dimensionality.
  • Effective dimensionality here means how many independent directions evolution has explored within a family, and many families show very few such paths.
  • Simulations suggest the first protein families likely needed DNA recombination to form rather than mutation from a single starting sequence.
  • The collaboration cautions that AI design tools learn from a tiny slice of functional proteins and calls for new datasets, noting support from Japan’s JST ASPIRE program and no competing interests.