The exceptional growth in new DNA sequences continues to provide equally exceptional opportunities for biologists to understand fundamental properties and principles of living cells. But it also presents a substantive barrier, forcing experimentalists to change their approaches to include a major computational role in experimental design itself. In particular, the rate of sequence acquisition in the case of complete genomes already exceeds the capacity of even the largest experimental lab or even the entire experimental community to investigate a significant number of the genes and demonstrate their function, which has meant that almost all information about gene function is obtained through computational predictions, and only a small fraction of those predictions are tested or validated experimentally. Yet, the predictions are used repeatedly for extensions, i.e., for creating annotations for newly sequenced genomes, leading to transitive errors and limiting the value of extant annotations and severely limiting progress. Accurate, generalizable approaches to filter experimental space by making functional predictions is just now being starting to be developed as an area of intense research. Computer science and information technology research is required to enable a deep understanding of contemporary biology as well as applications through the development of software tools that accurately can identify genes and predict their function. Such predictions need to be accurate enough that the experimental space is reduced, that the computational methods serve to filter the options and allow experimentalists to focus on the categories of new genes most likely to be useful. This annual meeting on the assessment of automatic function prediction will bring together the best research in an interactive environment to assess the utility of new tools and reveal the basic algorithmic advances that will promote rapid, accurate prediction of function of the many new genes being discovered.
