Ligand Similarity Complements Sequence, Physical Interaction, and Co-Expression for Gene Function Prediction
Likelihood that two proteins will be related by ligand similarity (solid line: SEA E-value < 1e-5, dashed line: SEA E-value < 1e-20) given a threshold in the (A) sequence similarity network, (B) co-expression network, and (C) extended protein-protein interaction network. The Y-axis is the likelihood that pairs of targets will have a SEA E-value better than 1e-5 (and, for sequence similarity, also 1e-20) at any given threshold of similarity on the X-axis. (D-F) Truth tables showing the correspondence of the protein-protein pairs that either are or are not related by ligand similarity and by sequence similarity, co-expression, or direct protein-protein interactions. In the upper left and lower right squares, the ligand-based and genomics association agree that the targets are or are not related, while in the lower left and upper right they disagree.