Improved Singular Value Decomposition-based Exons Prediction Approach Using Forward-backward Filtering

摘要

Exons prediction is a significant task in genomic signal processing to identify the locations of protein-coding regions in deoxyribonucleic acid (DNA) sequences. Singular value decomposition (SVD) is one of the approaches previously proposed for this purpose. This paper proposes a modified SVD-based exons prediction approach in order to improve the exons prediction accuracy. The improvement is achieved by employing the forward-backward filtering technique in the preprocessing stage to remove the non-linear phase distortion caused by the anti-notch infinite impulse response (IIR) filter. In order to test the exons prediction performance of the proposed approach against that of the traditional approach, MATLAB simulation is carried out on the ASP67 dataset. The receiver operating characteristics (ROC) and precision-recall curves in addition to the F1 score are utilized as performance evaluation metrics. The obtained results confirmed the merit of the proposed approach through achieving an improvement of 16.78% in the exons prediction accuracy at 10% false positive rate (FPR), compared to the traditional approach. In addition, the F1 score is increased by 5.16%.