Calls for new reporting standards, quality control in microarrays.

C oncerns about the quality of microarray data in three Duke University cancer trials have simmered for several years. Then in July, after evidence surfaced regarding alleged inconsistencies in reporting of credentials by one of the Duke researchers, the issue came to a boil: More than 30 bioinformatics and biostatistics specialists sent a letter to National Cancer Institute director Harold Varmus, M.D., and others expressing concern about the reproducibility of the data and asking that the trials be suspended. Within 3 days, they were. Now, using the momentum from that incident, the group of biostatisticians and bioinformatics specialists is calling for improved reproducibility and transparency in genomics experiments. Writing under the designation “Scientists for Reproducible Research Working Group,” they have sent a proposal to the editor of a major journal recommending that journals require microarray study authors to furnish more detailed methods to allow reproducibility. The group’s proposal is in line with data from the Microarray Quality Control II (MAQC-II) study, published online in Nature Biotechnology in July. The fi ndings from MAQC-II demonstrate that, although one can derive predictive models from microarray data, they need to be carefully developed and independently tested — and that reproducibility requires substantial effort.