Statistical DNA Sequence Modeling and Exon Detection Using Non-Parametric Methods

Abstract

This paper presents a hybrid approach based on digital bandpass filtering with non-parametric estimation techniques for the analysis of deoxyribonucleic acid (DNA) sequences. These spectral estimation techniques improve the analysis of DNA sequences and enable the extraction of some desirable information about them. The electron-ion interaction pseudopotential (EIIP) numerical representation method is used to convert a DNA sequence to numerical values through a mapping function. Also, mathematical modelling is used to create closed formulas for the represented DNA data sequences with different studied methods. The importance of this process is that the mathematical models can be used for any further processing or identification when applied to DNA sequences. The metrics used for performance evaluation are root mean square error (RMSE) and correlation coefficient (R) metrics. Also, the objective of this paper is investigating and predicting the location of the coding region (exon) in DNA sequences using the proposed approach. The results of gene prediction from DNA sequences for the original and modelled DNA sequences coincide and ensure the success of the proposed sum-of-sinusoids method for modelling of DNA sequences.