Semantic-Enhanced Classification of DNA Sequences Using Machine Learning and Deep Learning Approaches

The classification of DNA sequences is a core operation in genomics that helps provide semantic insight into genetic variation, evolutionary patterns, and biological roles. This study compares machine learning and deep learning models for alignment-free DNA sequence classification, focusing on semantic meaning of genomic patterns and result interpretability. Logistic regression (LR), support vector machine (SVM), multilayer perceptron (MLP), and convolutional neural network (CNN) were evaluated with 6-mer feature encoding applied to the genomic data of humans, chimpanzees, and dogs. Experiments demonstrated that MLP was the most accurate (94.99% versus 94.97%) and LR was the second (93.21%). CNN and SVM came in third and fourth, respectively (92.74% and 91.94%). In addition, unsupervised clustering with Principal Component Analysis (PCA) and t-distributed Stochastic Neighbor Embedding (t-SNE) can show semantically sensible groupings on the species level and, therefore, k-mer representations can maintain biologically significant structure. This study identifies trade-offs between predictive accuracy, computational efficiency, and semantic interpretability, with classical models providing transparent knowledge extraction and deep models involving complex hierarchical genomic relations. This work bridges data-driven learning and understandable genomic representations, thus helping to build semantic-aware bioinformatics analysis and aiding the integration of intelligent genomic systems into the broader paradigm of the Semantic Web and information systems.