Sir

Your Opinion article about gene expression, “Free and public expression” (Nature 410, 851; 2001), and the accompanying News Feature about DNA microarrays, “When the chips are down” (Nature 410, 860–861; 2001), reveal the tip of the iceberg.

Even larger problems loom ahead than the microarray data acquisition discussed in your articles. There is as yet no analytical standard to interpret the data. Traditional statistical algorithms are not sufficient for modelling functional genomics, and additional analytical problems abound, for example the issue of decoding the data structures that relate gene expression to gene function in complex biological systems.

We are undertaking a community-wide experiment to review state-of-the-art algorithms in microarray data analysis, to identify key research problems to be addressed. Last October, the Duke University Bioinformatics Shared Resource initiated a critical assessment of microarray data-analysis techniques (CAMDA) which invited the whole scientific community to analyse the same standard data sets.

CAMDA is a functional-genomics successor to other community-wide experiments, such as GASP (http://www.fruitfly.org/GASP1) in genomics and CASP (http://predictioncenter.llnl.gov) in protein modelling.

CAMDA 2000 mobilized researchers in traditionally separate fields of study, creating an interactive forum for a review of microarray-analysis techniques. For the first time, molecular biologists, statisticians, mathematicians, computer scientists, engineers and physicists were all involved in looking at a problem whose ultimate solution will require their collaboration. The plan is to have a meeting every year to evaluate and stimulate academic research and commercial developments in microarray data analysis (see http://www.bioinformatics.duke.edu/camda).