Analysis of DNA Sequence Classification Using CNN and Hybrid Models.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• DNA sequence classification is vital for biomedical data analysis and virus identification.
• Traditional methods face challenges in feature selection and high dimensionality.
• Deep learning offers automated feature extraction capabilities.

Purpose of the Study:

• To evaluate deep learning models for DNA sequence classification.
• To compare different architectures (CNN, CNN-LSTM, CNN-Bidirectional LSTM) and encoding methods (Label, K-mer).
• To improve accuracy in identifying viral DNA sequences.

Main Methods:

• Employed Convolutional Neural Network (CNN), CNN-LSTM, and CNN-Bidirectional LSTM architectures.
• Utilized Label and K-mer encoding for DNA sequence representation.
• Evaluated model performance using various classification metrics on testing data.

Main Results:

• CNN and CNN-Bidirectional LSTM models with K-mer encoding demonstrated high accuracy.
• Achieved 93.16% accuracy with CNN and 93.13% with CNN-Bidirectional LSTM on testing data.
• Deep learning models effectively extracted features from DNA sequences.

Conclusions:

• CNN and CNN-Bidirectional LSTM with K-mer encoding are effective for DNA sequence classification.
• These models show promise for virus identification, outbreak detection, and drug design.
• Automated feature extraction by deep learning addresses limitations of traditional methods.