Performance Evaluation of Different Window Functions for STDFT Based Exon Prediction Technique Taking Paired Numeric Mapping Scheme

During the last two decades, significant progresses have been made on advancing potential DNA sequencing techniques that generate huge amount of genomic data. Consequently, a great deal of research is being focused on investigating these data using mathematical and computational tools. Short time discrete Fourier transform (STDFT) is often used for spectral analysis of genomic signals that further help in retrieving useful information of molecular biology such as locating protein coding regions i.e. exons in a eukaryotic genome. However, identification accuracy of STDFT based exon prediction technique is highly influenced by the background noise present in its spectrum which sometimes masks the distinctive period-3 feature of protein-coding regions. Selection of proper window function play vital role in background noise reduction and therefore it is important to evaluate the performance of different window functions for exon predictions. Earlier, Voss mapping is used by researchers to test the window performances which is computationally costly and less informative. In this paper, performance of five important window functions is tested on different genes using paired numeric (PN) mapping scheme which is comparatively more advanced and computationally efficient than Voss mapping. Results are compared in terms of two parameters i.e. signal to noise ratio (SNR) and area under the ROC curve (AUC). It is found that the use of the Blackman window provides better prediction performance than other window functions with paired numeric mapping scheme. MATLAB 2013a is used for the simulation process.