LedPred: Learning from DNA to Predict enhancers

2 Description 2 2.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.2 Learning from CRM-contained information to predict new regulatory features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.2.1 Building the training set . . . . . . . . . . . . . . . . . . . 2 2.2.2 Optimization of support vector machine model . . . . . . 3 2.2.3 Definition of the optimal SVM parameters (γ and C) . . . 3 2.2.4 Sorting and selecting features according to their importance in the traing set description . . . . . . . . . . . . . 4 2.2.5 Plotting the performances of the model . . . . . . . . . . 4 2.3 Function description . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.3.1 From bed CRM genomic coordinates to training set matrix 5 2.3.2 Scaling of the CRM feature matrix . . . . . . . . . . . . . 5 2.3.3 SVM parameter optimization . . . . . . . . . . . . . . . . 5 2.3.4 Features ranking . . . . . . . . . . . . . . . . . . . . . . . 6 2.3.5 Selecting features . . . . . . . . . . . . . . . . . . . . . . . 7 2.3.6 Creating the best model . . . . . . . . . . . . . . . . . . . 7 2.3.7 Plotting model perfomance . . . . . . . . . . . . . . . . . 8 2.3.8 Using the model to score unknown sequences . . . . . . . 8 2.3.9 From the matrix to the model in one function . . . . . . . 9