Table 1.
Comparison of designs and results of CRISPR/Cas9 editing outcome prediction.
| Model | Cell Line(s) | Indels Predicted 1 | Methods 2 |
|---|---|---|---|
| Apindel | K562 | 536 classes of Deletions, 21 classes of Insertions. |
GloVe + Positional Encoding BiLSTM + Attention |
| CROTON | K562 | Deletion frequency,1 bp Insertion/Deletion, 1/2 bp Frameshift frequency, Frameshift frequency. |
CNN + NAS |
| Lindel | HEK293T | 536 classes of Deletions, 21 classes of Insertions. |
Logistic Regression |
| SPROUT | T cell | 9 types statistics of the repair outcomes such as average insertion length. |
Gradient Boosting Decision Tree |
| FORECasT | K562, RPE1, iPSC, CHOHAP1, mESCs |
~420 classes of Deletions, 20 classes of Insertions. |
Multi-Class Logistic Regression |
| InDelphi | HEK293, K562, HCT116, mESCs, U2OS |
~90 classes of MH Deletion, 59 classes of Non-MH Deletion, 4 classes of 1 bp Insertion. |
Deep neural network k-Nearest Neighbor |
1 Indels predicted indicates the repair labels that can be predicted by the model. 2 Methods represents the main algorithms used by the model.