Protein structure, the three-dimensional folding of a string of amino acids, determines protein function. Thus a knowledge of protein structure is essential to rational drug design, and manipulating protein structure could allow the generation of new or altered functions. However, the ability to predict structure from a primary amino acid sequence remains an elusive Holy Grail. X-ray crystallography currently is the only sure-fire way to elucidate a protein's structure, though there are less time-consuming methods such as circular dichroism and nuclear magnetic resonance that give an idea of what a protein looks like.

Now researchers at the California Institute of Technology have published a computer-based screening approach that identifies a sequence of amino acids predicted to conform to a specified structural motif. The researchers took a known 28-amino acid structure called a zinc finger - a common motif in DNA-binding proteins - and used a computer algorithm to pick the best amino acid sequence that would produce the desired conformation out of a virtual library containing 1.9X(1027) possibilities. As a comparison with combinatorial peptide synthesis, a library of this size with one polypeptide per possible sequence would weigh 11.6 metric tons.