Table 7.
Categories/models | Efficient features | Inefficient features |
---|---|---|
Overall nucleotide usage | ||
DeepCRISPR | A, U in the middle | C in the middle |
CNN-SVR | A, U in the middle | G count |
C-RNNCrispr | NN in positions 17–19 | |
Position-specific nucleotides | ||
DeepCRISPR | G in position 20 | A in position 20 |
C in positions 17–20 | U in positions 17–20 | |
G in position 16 | ||
CNN-SVR | A in position 20 | G in position 20 |
C in positions 17–20 | U in positions 17–20 | |
C-RNNCrispr | A in position 20 | A in positions 3–6 |
C in positions 17–19 | ||
U in position 20 | ||
Motifs | ||
DeepCRISPR | C in PAM (CGG) | T in PAM (TGG) |
G in PAM (GGG) | A in PAM (AGG) | |
CNN-SVR | T in PAM (TGG) | C in PAM (CGG) |
A in PAM (AGG) | G in PAM (GGG) | |
Epigenetic | ||
DeepCRISPR, CNN-SVR, | Open chromatin (esp. position 17) | Methylated DNA [RRBS] |
C-RNNCrispr | [CTCF, DNase] |
CTCF: CCCTC-binding factor, RRBS: reduced representation bisulfite sequencing, N: any nucleotide.
Conflicting results are marked in bold.