Table 2 Cross-validation results on the CIRCLE-seq dataset
From: A versatile CRISPR/Cas9 system off-target prediction tool using language model
Model | Bal Acc | F1-score | AUROC | AUPRC |
|---|---|---|---|---|
CCLMoff | 0.998 ± 0.001 | 0.409 ± 0.003 | 0.985 ± 0.001 | 0.524 ± 0.004 |
LSTM | 0.843 ± 0.002 | 0.052 ± 0.001 | 0.926 ± 0.001 | 0.479 ± 0.003 |
CRISPR-Net | 0.806 ± 0.002 | 0.083 ± 0.001 | 0.915 ± 0.003 | 0.462 ± 0.007 |
CCTop | 0.887 ± 0.004 | 0.003 ± 0.001 | 0.711 ± 0.004 | 0.008 ± 0.001 |
CCLMoff-Epi | 0.998 ± 0.001 | 0.429 ± 0.005 | 0.989 ± 0.001 | 0.513 ± 0.004 |
CCLMoff-Van | 0.836 ± 0.001 | 0.053 ± 0.001 | 0.901 ± 0.003 | 0.422 ± 0.005 |
CCLMoff v.s. | ||||
LSTM | 6 × 10−12 | 3 × 10−12 | 9 × 10−9 | 4 × 10−11 |
CRISPR-Net | 1 × 10−9 | 8 × 10−7 | 1 × 10−5 | 7 × 10−6 |
CCTop | 1 × 10−8 | 1 × 10−23 | 1 × 10−13 | 1 × 10−18 |
CCLMoff-Epi | 0.82 | 0.53 | 0.09 | 0.12 |
CCLMoff-Van | 1 × 10−4 | 2 × 10−13 | 6 × 10−4 | 5 × 10−6 |
