A NEW technique for predicting the functions of proteins may be a godsend for
scientists trying to make sense of human genome maps due to be completed early
next century.
A gene鈥檚 DNA sequence reveals which amino acids will appear in the protein
the gene encodes. To guess a new protein鈥檚 function, scientists often compare
its amino acids with those of familiar proteins. If two proteins share much of
their sequences, the chances are they do the same job.
But this takes no account of the protein鈥檚 shape, so the predictions are
often wrong. 鈥淲hat makes a protein function is its three-dimensional structure,鈥�
says Jacquelyn Fetrow of the Scripps Research Institute in La Jolla, California.
Advertisement
Fetrow and her colleagues Jeffrey Skolnick and Adam Godzik can now forecast a
protein鈥檚 function more accurately by guessing its 3D structure. They use a
computer program that predicts which of 300 known protein shapes the unknown
protein can fold into, and assume the actual structure is the most stable of
these. They then pick out amino acids that play important roles in other
proteins, and decide where on this structure they lie. This reveals the active
sites, giving a good indication of what the proteins do (Journal of
Molecular Biology, vol 282, p 703).
The team tested the method on proteins that the Escherichia coli
genome encodes. They identified the 10 E. coli proteins that belong to
a class of enzymes called thiol-disulphide oxidoreductases. 鈥淚t鈥檚 a major
advance,鈥� says Stanley Krystek of pharmaceuticals company Bristol-Myers Squibb
in Princeton, New Jersey.