Strong association between genomic 3D structure and CRISPR cleavage efficiency
Fig 3
3D features significantly improve CRISPR prediction.
Histograms of the correlations between measured and predicted efficiency in the Leenay and TTISS datasets, when testing LASSO and xgboost models. The blue/orange histogram indicates the model with/without the 3D feature and the average correlation is marked with a solid/dashed line, respectively. p-values were calculated using Wilcoxon’s signed rank test. (A) Models using all 425 classic features; (B) Models using the top 30 features, based on Pearson correlation with CRISPR efficiency.