Enhanced DNA Sequence Classification Through Hyperparameter Optimized Convolutional Neural Networks

DNA sequence classification is a fundamental task in bioinformatics, with applications ranging from gene prediction to disease diagnosis. Convolutional Neural Networks (CNNs) are a deep learning-based method that is suggested for DNA sequence classification. Two significant biological datasets, Promoter and Splice, are utilized for evaluation. The raw DNA sequences are transformed into numerical representations through k-mer feature extraction, serving as input for the CNN model. Multiple convolutional and pooling layers, followed by a fully linked layer for classification, make up the carefully planned CNN architecture. Extensive hyperparameter tuning is conducted to optimize the model, including adjustments to parameters such as learning rate, kernel size, and number of layers. Experimental results indicate that the tuned CNN model achieves a classification accuracy of 97% on both the Promoter and Splice datasets. The performance of the proposed model is compared with other baseline machine learning models, demonstrating its superior effectiveness. This study underscores the utility of CNNs in DNA sequence classification tasks and provides valuable insights for genomic research and applications in bioinformatics.