Figure 3From: Evaluating a variety of text-mined features for automatic protein function prediction with GOstruct Functional class analysis of all GO term annotations and predictions. a) Distribution of the depth and information content of GO term annotations. As IC values are real numbers, they are binned, and each bar represents a range, e.g. ‘[1,2)’ includes all depth 1 terms and IC between 1 and 2 (not including 2). b) Macro-averaged F-measure performance broken down by GO term depth. c) Macro-averaged F-measure performance binned by GO term information content.Back to article page