Figure 1
From: Predicting gene ontology functions from protein's regional surface structures
Similar pockets imply similar functions. The relationship between pocket similarity and functional similarity. The graph summarizes the results in Tables 1-4. We use four statistical significance measurements as the thresholds to construct the similarity network, that is E-value of the sequence similarity, p-value of structure similarity cRMSD, p-value of structure similarity oRMSD, and the combination of E-value and p-value of cRMSD. The horizontal axis represents the number of GO annotated edges (i.e. pocket pairs) found by given thresholds. The vertical axis denotes the percentage of GO annotated edges with at least one identical GO term between two end nodes. The graph shows that the structure-based threshold is better than the sequence-based threshold.