Replication Publication

I n December 2003, PLoS Biology published a research article reporting evidence of an association between a gene called GAD2 and susceptibility to obesity (DOI: 10.1371/ journal.pbio.0000068). Although the genetic data were suggestive rather than conclusive, the editors and reviewers supported publication because unambiguous evidence in association studies is notoriously diffi cult to obtain and obesity is such a major burden on public health. Enthusiasm was heightened because GAD2 lies in a region of human Chromosome 10 that is thought to contain a gene infl uencing obesity, and GAD2 itself is involved in the synthesis of a neurotransmitter implicated in the regulation of food intake. Publication would allow others to test just how important the connection between GAD2 and obesity might be—and the enticing prospect was that insight into the role of GAD2 in this disease might lead ultimately to new therapeutic approaches. In this issue of PLoS Biology, we publish a follow-up to the original study, from a separate team of researchers based at the University of California at San Francisco (DOI: 10.1371/journal. pbio.0030315). Swarbrick et al. looked at different and much larger patient populations and analyzed variation within GAD2 more comprehensively, but disappointingly, the new study was not able to replicate the initial fi nding. Despite the negative conclusions, however, the reviewers were no less enthusiastic about the publication of this paper than the fi rst. One of the reasons for this is related to a problem that has beset the fi eld of complex disease genetics for several years—a tendency towards the publication of studies that show an association. Diseases are termed complex when their etiology is infl uenced by many factors, both genetic and environmental. These diseases include obesity, diabetes, mental illness, and many more common ailments. But fi nding the genes that are involved with these diseases is an extremely diffi cult problem: individual genes have only a limited effect, which means that large patient populations need to be studied; specifi c genes might have effects only in certain ethnic groups; and statistical artifacts can be hard to eliminate. These and other problems mean that success stories are few and far between, whereas false leads have been plentiful. High-profi le journals, in particular, have therefore tended to publish the positive results, whereas negative data often end up in specialist literature or, much worse, don't get published at all. Because the identifi cation of a complex …