SVM Helitrons recognition based on features extracted from the FCGS representation

Abstract

It is well recognized that the signal processing methods contributes in biology to the control of the DNA spatial structure. From the previous studies, it is inferred that the significant portion of the eukaryotic genomes is composed of transposable elements (TEs). The TEs play an important role as a driving force of genome evolution. An important sub class of ETs class II, Helitrons, have been revealed in diverse eukaryotic genomes. These elements have a remarkable ability to capture genes and they transpose by a rolling circle mechanism. In this context, helitron location is one of the main challenges of cell biology to better understand the different hereditary characteristics transmission modes in genomes. In this paper, we introduce the Frequency Chaos Game Signal (FCGS) method which provides a new way to present the DNA genomic sequence as a DNA signal (1-D presentation). Then, we use the Support Vector Machine (SVM) classification technique to classify helitron DNA. This choice came after studying the performance of SVM technique by varying the parameters of the kernel tricks. For this aim, different kernels have been taken into account. As for the nonlinear SVM approach, we choose the one-against-one strategy.