Fig 1.
Schematic diagram of the proposed hybrid fuzzy model.
The proposed model introduces two main components: an information theory-based approach and a fuzzy model. In the fuzzy model, the dashed box indicates the steps by which the regulatory genes are selected.
Fig 2.
Fuzzy membership function of the regulatory effect of repressors and activators.
It has three discrete levels: Low, Medium, and High.
Fig 3.
Fuzzy membership function of target gene expression level.
There are six discrete levels: VL = Very Low, L = Low, Med = Medium, H = High, and VH = Very High.
Fig 4.
Fuzzy decision rule matrix to predict the expression level of a target gene based on the regulatory effect of activator and repressor gene pairs.
Table 1.
Comparison of True Positive Rate, False Positive Rate, F-score, Matthews correlation coefficient, and Structural Accuracy produced by the Information theory-based approach, and MICFuzzy with and without the inclusion of regulatory relationship strength.
Fig 5.
Comparison of Average F-score, Average MCC and Average Structural Accuracy of MICRAT, NARROMI, and MICFuzzy for inferring DREAM4 10-gene networks.
Results of (a) Average F-score (b) Average MCC and (c) Average Structural Accuracy.
Fig 6.
Comparison of Average F-score, Average MCC and Average Structural Accuracy of MICRAT, NARROMI, and MICFuzzy for inferring DREAM4 100-gene networks.
Results of (a) Average F-score (b) Average MCC and (c) Average Structural Accuracy.
Table 2.
Comparison of proposed model with other fuzzy models for inferring DREAM3 and DREAM4 networks.
Fig 7.
Reduction of combinatorial computation of the MICFuzzy model compared to the combinatorial computation required for the classical model in inferring DREAM networks.
Inferring results of (a) DREAM3 10-gene, (b) DREAM3 50-gene (c) DREAM4 10-gene and (d) DREAM4 100-gene networks.
Fig 8.
The bacterial E. coli SOS DNA repair target network.
Fig 9.
Structure of SOS DNA repair network reconstructed by MICFuzzy.
(a) Target network with all true regulations. Black lines indicate true regulations which are not predicted by MICFuzzy and blue lines indicate regulations correctly inferred by MICFuzzy with identified regulation type. (b) Both true and false regulations are inferred by MICFuzzy. Blue lines indicate regulations correctly inferred by MICFuzzy and red dashed lines indicate false predictions. In this figure, arrows and barred lines represent positive and negative interactions respectively.
Table 3.
Comparison of true regulations inferred by MICFuzzy with other methods in the SOS DNA repair network.