Figure 3

Comparison of the performance of a single SVM with other methods in predicting BP annotations of the novel set. This figure shows the AUC achieved by different submissions to the mouse function prediction project compared to a single SVM approach for GO BP terms of different sizes: (a) 3 to 10; (b) 11 to 30; (c) 31 to 100; and (d) 101 to 300. We found that a single SVM, after careful adjustment of parameters, performed relatively well across different sizes of GO terms, indicating the robustness of the SVM as a baseline prediction method. Only groups C, D and G performed better, on average, than the single SVM results. The dashed red lines represent the performance of single SVM in terms of AUC. Asterisks indicate second round submissions. AUC, area under receiver operating characteristic curve; BP, biological process; GO, Gene Ontology; SVM, support vector machine.