Sir

We were delighted by your News report “Computer modellers seek out 'Ten Most Wanted' proteins” (Nature 409, 4; 2001), highlighting recent advances in protein structure prediction.

The report states, however, that computational structural biologists “believe that computer modelling may become a viable alternative to the current practice of using X-ray crystallography and nuclear magnetic resonance techniques to 'solve' the structure of a protein”.

The importance and value of structure prediction methods is to produce rough, approximate models for proteins with no experimental structure available. These models can often provide verifiable hypotheses about function, as your report accurately mentions. However, in the forseeable future we expect predicted structures neither to approach the accuracy of experimentally determined structures, nor to replace them.

The report also states: “The big question, researchers say, is whether modelling will continue to improve fast enough to make an impact. There is a relatively short window of time for structure prediction to be useful before structural genomics . . . generates all the results needed.”

We disagree strongly with this statement. Structural genomics aims not to solve the structures of all proteins, but rather to obtain a set of representative structures such that all others can be modelled. Thus computational methods will play a critical role in translating the information on the relatively small fraction of proteins whose structures will be solved into accurate models for all proteins.

We furthermore expect that computational methods will produce useful low-resolution models for large numbers of sequence families many years before representative structures can be solved experimentally, and will play an all-important role in the design of novel proteins and therapeutics.