Fig 13. Instance-based explanation (enhancer experiment, layer 1).
Instance-based explanation of one sample of each type, strong enhancers (SE), weak enhancers (WE), and non-enhancers (NE). Due to length of the sequences, only relevant parts of the instanced-based explanation are shown. We can observe that there are multiple relevant motifs, which also depend on the enhancer type (WE or SE).