Fig. 1From: Functional networks inference from rule-based machine learning modelsTwo approaches to functional network inference: one based on the expression profile similarity and the other based on the extraction of knowledge from machine learning models. The similarity-based methods construct a new network edge X⇔Y, when the similarity between the expressions of genes X and Y across the samples is above a threshold. Methods based on machine learning, first build a predictive model, in this example a rule-based model, using the samples phenotype (class labels) information and then construct a network edge X⇔Y, when genes X and Y are used together within that model to classify the samples. As these two approaches lead to different functional networks, it is possible that they capture complementary knowledgeBack to article page