A novel method of recognizing short coding sequences of human genes

In this paper, we present a novel feature representation of DNA sequences based on the graphical representation. Support vector machine (SVM) is applied to classify the coding/non-coding sequence in short human genes. In the process of identifying, we propose an improved self-similar map method to avoid the lack of negative samples sequence. According to the GC content we divide the dataset into several groups and identify these sequences respectively. Finally, the results show that the proposed method obtains a higher accuracy with fewer parameters.