Figure 6
From: Learning Interpretable SVMs for Biological Sequence Classification
We use Multiple Kernel Learning to determine weights for the WD kernel. Shown is learned weighting for the WD kernel at the acceptor and at the donor site. From areas of higher weight (upstream intron: regions -70 nt to -40 nt and -30 nt to 0 nt, Exon: +30 nt to +70 nt and -30 nt to -90 nt, downstream intron 0 nt to +70 nt) overrepresented hexamers have been extracted and are shown in Table 1.