Fig. 3

Performance comparison on simulated benchmarking datasets. a Simulated datasets containing true samples with highly expressed miRNA genes in sender cells and lowly expressed target genes in receiver cells for negatively regulated MiTIs. Boxplots displayed the distribution of AUROC and AUPRC across simulated datasets for each method. For the boxplots showing the minima, 25 th percentile, median, 75 th percentile, and maxima, the number of data points for each method is 20 with the median value shown beside each box. b Simulated datasets for the inference of autocrine communications mediated by MiTIs. Point plots displayed Pearson correlation coefficients and corresponding P values of inferred MiTIs by miRTalk on the simulated dataset containing 1000 cells and 1000 gens with 50% miRNA/target coverage with an example of inferred MiTI, namely miRNA346-target180 interaction. c The Pearson correlation coefficients between the gene expression of miRNAs and their targets on the simulated datasets and the ratio of significant MiTIs with negative regulation inferred by miRTalk and CSmiR. For the boxplots showing the minima, 25 th percentile, median, 75 th percentile, and maxima, the numbers of data points from left to right are 500, 709, 486, 722, 478, 722, 490, 750, 460, 728, 484, 772, 488, 755, 482, 729, 503, 675, 466, 707, 500, 709, 326, 475, 180, 262, 77, 110, 18, 31, 500, 709, 324, 463, 185, 263, 77, 128, 14, and 30 respectively. AUROC, area under the receiver operating characteristic curve. AUPRC, area under the precision-recall curve