We have attempted to make our interface and algorithm flexible enough that arbitrary gene networks can be uploaded. Although we depend heavily on annotation and curation efforts of others, we try to avoid doing curation or make choices for the user ourselves. This means that by default, we include all network data that we have for the organism with two exceptions: we currently include only the 20 co-expression data sets that are generally most informative for gene function prediction to speed up queries that use the default settings; also, we do not include functional interaction networks generated using Gene Ontology (GO) annotation because they introduce circularity in our calibration and performance measuring algorithms, which currently rely on performing cross-validation with GO annotation. Nonetheless, both of these network types can be added to the query using the advanced options panel.
