Figure 2.

Flow scheme of model building To improve model interpretability and reduce overfitting, sophisticated computational strategies implement feature selection algorithm to select a subset of relevant features for model building. Then, appropriate classification model is employed to differentiate active enhancers from non-enhancers. Generally, there are two major classification models. The first is the discriminative models which find the optimal classification border in the feature space (lower left panel). The other one is the probabilistic graphical models that try to model the joint distribution of states and associated features with graph (lower right panel). ANN, artificial neutral network; BN, Bayesian network; HMM, hidden Markov model; SVM, support vector machine.