Fig. 2
From: Iterative sub-network component analysis enables reconstruction of large scale genetic networks
Synthetic, toy case-study and comparison between ISNCA and standard NCA. a The complete network is composed of four target genes (TGi, gray) and two transcription factors (TFi, red). The two sub-networks of the ISNCA are encircled (solid line) and share TG1 and TG2. b Comparison of the error between the ISNCA (white) and the GNCA-r (red). The mean error of 100 simulations (see ‘Methods’) of the ISNCA was significantly lower (p < 10−12, Kruskal-Wallis test) than of the GNCA-r. The error bars are the standard error of means (SEM). c The distribution of the error of the ISNCA (blue) and the GNCA-r (red). The errors of 91 % of the ISNCA simulations were under 0.04, compared to only 50 % of the GNCA-r. d The convergence of the ISNCA mean error (n = 100) was rapid, stabilized after 3–5 iterations. Shaded area represents standard deviation at iterations