Figure 3
From: Homology Induction: the use of machine learning to improve sequence similarity searches
This figure shows the calculated areas under ROC curve for both HI methods (HIall and HIseq) for a range of re-sorting factors. The AUROC values for HIall increases steadily and reaches its maximum value at 6 × 10-5 with a value of 0.651; while the AUROC values for HIseq first increases and then decreases again with a peak at 8 × 10-2 with an AUROC value of 0.613. Comparing both methods with PSI-BLAST shows that HIseq has a smaller improvement over PSI-BLAST which has an AUROC value of 0.607. HIall increases the AUROC value by approximately 7.4 per cent.