Fractal Analysis of DNA by Nonlinear Genome Signal Processing for Exon and Intron Separation

Abstract

Aims: To provide a new reasonable measure for distinguishing between coding and non-coding regions of DNA sequences based on its fractal nature and self-similarity. Study Design: After conducting background studies on the fractal structure of DNA sequences, the application of Detrended Fluctuation Analysis for identifying coding and non-coding regions in those sequences was investigated. Finally, the propositions were tested on a standard dataset of 195 genes. Place and Duration of Study: Sample: We use a common data set, “HMR 195”, which has been used in conventional tools, between December 2012 and July 2013. Methodology: The Fractal Scaling Exponent (FSE) of the numerical signal, produced by converting a DNA string to a numerical sequence via a number mapping algorithm, was calculated for exons and introns of 195 genes. This calculation was repeated twice: once for computing the optimal values of FSE, and once for non-optimal FSEs. Analysis of Variance (ANOVA) was used for investigating the significance of difference between the average FSE of exons versus that of introns in both optimal and non-optimal cases. Results: ANOVA indicated a significant gap between the optimal mean FSE of exons (0.65) and introns (0.72). The difference, although smaller, was significant for non-optimal values as well. Conclusion: Throughout this study, the FSE is proved to be a reliable measure for distinguishing between coding and non-coding regions of DNA gene sequences based on our experiments. Accordingly, this metric can be used for predicting exons/introns when embedded within current tools such as TestCode. However, its contribution to the predictive accuracy of current methods requires more investigation in the future works.