Splicing sites prediction of human genome using machine learning techniques

• Published in: Multimedia tools and applications Vol. 80; no. 20; pp. 30439 - 30460
• Main Authors: Ullah, Waseem, Muhammad, Khan, Ul Haq, Ijaz, Ullah, Amin, Ullah Khattak, Saeed, Sajjad, Muhammad
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.08.2021; Springer Nature B.V
• Subjects: 1155T: Advanced machine learning algorithms for biomedical data and imaging; Accuracy; Algorithms; Breast cancer; Composition; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Deoxyribonucleic acid; DNA; Gene sequencing; Machine learning; Multimedia Information Systems; Mutation; Ribonucleic acid; RNA; Robustness (mathematics); Special Purpose and Application-Based Systems; Splicing; Support vector machines; Big data analysis; Splicing sites; Genomics; Biomedical data; Machine learning; Pattern recognition; Computer-aided diagnosis
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-021-10619-3

The accurate splice site prediction has several applications in the field of medical sciences and biochemistry. For instance, any mutation affecting the splice site will lead to genetic diseases and cancer such as Lynch syndrome and breast cancer. For this purpose, collecting the Ribonucleic Acid (RNA) samples is an efficient and convenient method to detect the involvement of splicing defects in disease formation. Therefore, the present study aims to develop an accurate and robust Computer-Aided Diagnosis (CAD) method for swift and precise targeting of splice site sequences. A composite features-based model is proposed by integrating three different sample representation methods i.e., Dinucleotide Composition (DNC), Trinucleotide Composition (TNC) and Tetranucleotide Composition (TetraNC) for precise splice site prediction after converting the DNA sequences into numerical descriptors. The precision and accuracy of these features are analyzed by applying different machine learning algorithms such as Support Vector Machine (SVM), K-Nearest Neighbor (KNN) and Naïve Bayes (NB). Results show that the proposed model of composite features vector with SVM classifier achieved an accuracy of 95.20% and 97.50% for donor and acceptor sites datasets, respectively.