PredicircRNATool Files

The program can be used to predict circular RNAs from those of non-circularized, expressed exons based on conformational and thermodynamic properties in the flanking introns. Firstly, we extracted different length of sequences in the flanking introns and calculated the thermodynamic and conformational dinucleotide properties as the original features. And then two feature selection methods, Minimum Redundancy and Maximum Relevance (mRMR) and Random Forest (RF), were used to generate the optimized feature subset. For predictions, the performance of three algorithms, Support Vector Machine (SVM), Artificial Neural Network (ANN) and RF were compared on the training dataset and the independent dataset, respectively. 

Introduction:
1.All the data used for feature selection, training process and testing process is stored in the 'Dataset' folder. 
2.Dependency: the program can run on the Linux platform co-operating with MATLAB. For MATLAB, the libsvm tool ( http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/) should be installed.
3.For predictions, just run the 'circRNA_prediction.m' file in the model file. 50 nt sequences extracted in the flanking introns were combined as the input data.
4.The program will create a 'results.txt' file to display the results.