The LC dataset augments functional predictions. (a) Evaluation of curated literature against GO biological process as a standard. Comparisons of enrichment for functional relationships in LC dataset versus a variety of HTP datasets as scored against GO biological process are shown as the individual data points. The effect of the LC dataset on the predictive power of a Bayesian heterogeneous integration scheme  is shown by the curves. FN, false negatives; FP, false positives; TP, true positives. (b) Comparison of functional diversity in LC versus a variety of HTP datasets. The number of distinct functional groups (GO biological process terms) spanned by the LC dataset at decreasing levels of precision and recall. One hundred and forty-six independent GO terms were tested, all with fewer than 300 total annotations. A minimum F-score threshold (harmonic mean of precision and recall) was plotted against the number of GO terms needed to achieve that threshold for each of the data types.
Reguly et al. Journal of Biology 2006 5:11 doi:10.1186/jbiol36