Efficient Classification of DNA Microsatellites Using a Hybrid CNN-RNN Model
This research developed an efficient hybrid convolutional and recurrent neural network architecture that achieves high accuracy for classifying simple sequence repeats (SSRs) in DNA sequences with remarkably few training epochs. SSRs, also known as microsatellites, are short tandem repeating DNA motifs with broad utility as genetic markers. The model uniquely integrates convolutional neural networks (CNNs) and bidirectional long short-term memory (LSTM) recurrent layers to extract spatial and sequential features from sequence data. An attention mechanism focuses the model on the most informative regions. When tested on SSR sequences from the cricket Gryllus bimaculatus, this CNN-RNN framework attained 99.40% accuracy after only 5 training epochs. By blending complementary CNN and RNN methodologies, this work pushes the boundaries of efficiency for DNA sequence analysis. The highly flexible neural architecture establishes a foundation for rapidly advancing sequence classification accuracy with minimal training requirements. Overall, the model provides unparalleled accuracy on challenging genomics tasks using a fraction of previous training epochs