Abstract
Conference Title: 2017 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) Conference Start Date: 2017, Dec. 12 Conference End Date: 2017, Dec. 15 Conference Location: Kuala Lumpur, Malaysia The so-called DNA period-3 signals are significant indicators for exons locations in DNA sequences. However, the sole dependence on one period-3 signal to identify exons locations in a DNA sequence provides limited performance. It has been recently shown that exons prediction performance can be improved by employing multiple different period-3 signals. In this paper, we aim to justify the reason of this improvement by investigating the correlation between different DNA period- 3 signals. For this purpose, we utilize Pearson's correlation coefficient. Moreover, we examine another configuration of employing multiple period-3 signals for exons prediction that is not handled in previous work. MATLAB simulation is conducted on the HMR195 genomic dataset utilizing the ROC curve as a performance evaluation tool. The results reveal an average weak correlation of 12% between period-3 signals obtained by using different numeric mapping schemes. Consequently, employing those period-3 signals simultaneously provides better exons prediction performance as compared to employing either of them solely. On the other hand, the results show an average strong correlation of 80% between period-3 signals obtained by using different period-3 detection methods. Thus, the participation of those period-3 signals simultaneously does not provide a significant improvement in exons prediction performance as they carry similar information.