DNA Splice Junction Prediction Using Hybrid Approach of LSTM-GRU Model

DNA splice junction classification plays a critical role in computational biology, aiming to predict junction types (IE, EI, or N) from a given DNA sequence. This task is essential for understanding gene expression, disease mechanisms, splicing regulation, and the overall gene structure. Identifying the precise locations where exons are joined and introns are removed during RNA splicing remains challenging due to the lack of a universal rule governing this process. Furthermore, the complexity of splice junctions is compounded by the vast variability in genetic sequences across different organisms. This study explored the use of Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) for classifying DNA sequences, as these models are particularly effective in capturing long-range dependencies and contextual relationships between distant nucleotides. By leveraging the ability of LSTM and GRU to model sequential data, this approach was well-suited for understanding the intricate patterns inherent in splice junctions. The analysis was conducted using the StatLog DNA dataset, and model performance was evaluated across several metrics: Accuracy (96.48%), F1-Score (97.52%), Kappa (94.95%), Recall (99.38%), Precision (96.12%), and Coverage (97.48%). The results demonstrate the effectiveness of LSTM and GRU in predicting splice junction types with high accuracy and reliability, offering promising insights for future research in gene splicing and related bioinformatics fields.