Fig. 2.

Segregating guide and gene effects produces a predictive model for CRISPRi guide efficiency. A An overview of the training process of the MERF model. The MERF model segregates depletion values into predictions from a fixed-effect random forest model capturing guide efficiency and a random-effect linear regression model capturing effects associated with the target gene and dataset. The trained fixed-effect random forest model is used for gRNA efficiency prediction and the web-based tool CIAO (ciao.helmholtz-hiri.de). B Evaluating predictions of guide efficiency after removing gene effects. Spearman correlations between predictions and measured logFC for held-out genes. Genes were held out in tenfold cross-validation, and the reported median Spearman correlation was calculated across all held-out genes