Fig. 5

Network inference results using ground truth data (without downsampling) from 100 different simulated scale-free networks with 20 genes for 3 different network sparsities using additive or multiplicative regulation. Each network was simulated over 500 hours using parameters sampled as described in Methods sect. "Parameters for mammalian cells" A and E show barplots of the AUROC score for the 4 different network inference algorithms considered as well as a random classifier (RAND) for additive and multiplicative regulation respectively. B and F show barplots of the AUPR score for the 4 different network inference algorithms considered as well as a RAND classifier for additive and multiplicative regulation respectively. Confidence intervals for barplots were computed by subsampling 35 out of 100 networks 100 times. C and G show boxplots of the true positives found for each network inference algorithm and random classifier for 3 different sparsity levels for additive and multiplicative regulation respectively. The horizontal lines depict the actual number of true positives for reference. D and H show boxplots of the true negatives found for each network inference algorithm and random classifier for 3 different sparsity levels for additive and multiplicative regulation respectively. Again, the horizontal lines depict the actual number of true negatives for reference. E and I show boxplots of the false positives found for each network inference algorithm and random classifier for 3 different sparsity levels for additive and multiplicative regulation respectively. F and J show boxplots of the false negatives found for each network inference algorithm and random classifier for 3 different sparsity levels for additive and multiplicative regulation respectively