Fig. 2

Average number of experimentally confirmed miRNA-mRNA correlations on three datasets. This bar-chart represents the average number of the experimentally confirmed miRNA-mRNA correlations of each method on EMT, MCC and BR datasets. It also shows the statistical significances of differences on performance between two methods in terms of the p-value computed using the paired t-test. ‘Single’ means the average performance of three models from the three direct correlation inference methods without bootstrapping and Ensemble aggregation steps. ‘Bootstrap’ means the average performance of the bootstrapping aggregation results for each three direct correlation inference method. ‘Ensemble’ means the average performance of inferred models using Ensemble aggregation of single experiments. Additionally, Bootstrap&Ensemble means the average performance of proposed DMirNet that uses both bootstrapping and Ensemble aggregation. E-P&I&L means a comparable control that is an ensemble model aggregating Pearson, IDA and Lasso [1]