Fig. 6

Choosing the substitution matrix for fold recognition. We compare the efficiency of FASTA pairwise sequence alignment method SSEARCH (black), with the weighted string kernel WSeqKernel method with four different input K ₁ matrices, (BL62) (red), Optim10 (cyan), MIQS (blue), and VTML (magenta) to detect fold similarity. “True” relationships are defined according to CATH topologies. The analysis is performed on CATH40e4 that contains 10619 sequences corresponding to 1363 folds (see Methods for details)