Classification of exon and intron regions obtained using digital signal processing techniques on the DNA genome sequencing with EfficientNetB7 architecture

Abstract

Graphical/Tabular Abstract Figure A. The success rate obtained as a result of 100 epochs in the classification scope performed with the EfficientNet B7 architecture Purpose: To provide a classification of exon and intron labelled images obtained using digital signal processing techniques on the DNA genome sequence. Theory and Methods: In this study, the gene sequences numbered EU447303.1, AM400881.1, AM886138.1, AM600680.1 with different nucleotide lengths obtained from the NCBI database was examined. Firstly, DNA sequences that consist of symbolic characters were digitized using numerical mapping techniques. Then, digitized DNA sequences were expressed in the frequency domain by the windowed Fourier transform method and regions tagged as exon and intron were classified by transfer learning architectures. EfficientNetB7 architecture that located of transfer learning architectures presented maximum success by hydrogen bond energy rule of paired mapping technique. Secondly, the DNA sequences digitized using the hydrogen bond energy rule of paired mapping technique were expressed with spectrograms labelled as exon and intron regions by the windowed short-term Fourier transform method. Subsequently, spectrograms classified by EfficientNetB7 architecture produced the most successful results with a %100 success rate on the BCR-ABL genes. Results: This article was conducted to differentiate exon and intron regions on the BCR-ABL gene obtained from the NCBI database. As a result of the classification carried out with EfficientNetB7 architecture, which is a transfer learning technique, the success rate was obtained as %100. Conclusion: A successful accuracy rate and a low loss value have been obtained in the scope of the studies performed on the BCR-ABL gene with numerical mapping techniques, digital signal processing methods and transfer learning techniques.