A piece of cake.

I n the past year or so, I have devoted a lot of time to compiling manuscripts that incorporate outputs from genome-wide screens or profiles, which has required a lot of manual inspection of database entries. While those who work in computational biology may groan and advise me to let data-mining tools do the work—and do it more accurately, comprehensively, and objectively—I feel I have learned a lot, despite the excessive time invested. Most importantly, I have been able to absorb, digest, and condense the data into concepts that I could both understand myself and communicate to others, at least in a manner sufficient to satisfy peer-reviewers. Most of us in molecular biology use the excellent resources of NCBI (the US National Center for Biotechnology Information) on a daily basis, courtesy of the US taxpayer, and partner organizations. Despite inevitable duplication and the constraints of a standardized annotation system, the information available from NCBI is generally clear and user-friendly. However, my overall experience of other publicly available databases of genomic, molecular, expression, and phenotypic information is less positive. Each of these resources tends to have its own virtues and idiosyncracies, often beloved by the community at which they are primarily aimed, despite obvious limitations. But many repositories seem to have some major, closely related deficiencies. The information they contain is typically over-inclusive and ‘flat’. In other words, there is little selection of what is really important or fundamental to each entry, with contradictory, outdated, anecdotal, and sometimes largely irrelevant data commonly placed alongside recent, systematic or experimentally rigorous findings. The naı̈ve reader is left bewildered and befuddled by a mass of unsifted information, and making sense of it often requires one to find and evaluate the primary literature. The simplest questions—what actually is this gene, protein, pathway, or structure, and what is its function—can take a lot of effort to answer. In many ways, this negates the very idea of having electronic compendia that we can rely upon. Consider the following example, admittedly a bit exaggerated, but which illustrates the problems encountered in trying to understand biological database entries.