Fig. 4From: Regulatory analysis of single cell multiome gene expression and chromatin accessibility data with scREGValidate the RE-TG prediction. A, B Receiver operating characteristic (ROC) curve and precision-recall (PR) curve by taking the eQTL data of CD14 positive monocyte cells as ground truth to validate the prediction of scREG. Curves were plotted by sliding the predicted cis- regulatory score of 100,000 peak-gene pairs. As an alternative method, the Pearson correlation coefficient (PCC) between the expression of a gene and the accessibility of peaks within 1 Mb of the gene’s transcriptional start site was calculated. We can see scREG prediction achieves 0.81 area under the ROC (AUROC) curve and 0.46 area under the PR (AUPR) curve, which are much higher than that from PCC (0.56 and 0.25 respectively). C Comparison of the precision of scREG with random selections. Ratio of consistency (precision) was the percentage of RE-TG pairs validated by eQTL for those REs that linked to at least one gene in eQTL. Red line represents the ratio of consistency of scREG, the blue distribution represents that from 1000 random selection of peak-gene pairs in 1 Mb distance, and the orange distribution represents that from 1000 random selection of peak-gene pairs but restricting the distribution of distance between peak to genes is the same as the scREG predictions. D Validation of RE-TG prediction by HiC data on Naïve CD4 T cell. All metrics are calculated same as in C but replacing the eQTL by HiCBack to article page