Fig. 3

Accurate prediction of chromatin loops using single cell data. A machine learning model was trained on aggregated scRNA-seq and scATAC-seq data in GM12878, using ChIA-PET and Hi-C data as ground truths. A Overview of the pipeline, which yields a random forest classifier and can be used to make predictions on a new dataset. B Feature importance scores = mean decrease impurity in the training process; “avg” and “std” represent the mean and standard deviation of the signal intensity on both anchors; “left” and “right” represent flanking features and “in-between” is the signal intensity within a loop. C Receiver operator characteristic (ROC) curve and D precision recall (PR) curve, using five iterations of tenfold cross validation