Splice-site identification for exon prediction using bidirectional LSTM-RNN approach

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate prediction of DNA and protein sequences is crucial for biological research.
• Splice-site identification in eukaryotic DNA is challenging due to high false positive rates.

Purpose of the Study:

• To develop a deep learning model for accurate splice-site identification and exon prediction in eukaryotic DNA.
• To address the limitations of existing methods in handling splice-site prediction.

Main Methods:

• A bidirectional Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) deep learning model was designed.
• The model utilizes intron features, including donor (GT) and acceptor (AG) sites, considering length constraints.
• Model performance was enhanced by increasing training epochs.

Main Results:

• The developed bidirectional LSTM-RNN model achieved a maximum accuracy of 95.5% in splice-site prediction.
• The model demonstrated effectiveness in identifying splice sites for exon prediction from eukaryotic DNA sequences.

Conclusions:

• The bidirectional LSTM-RNN model offers a robust solution for accurate splice-site identification in eukaryotic genomes.
• This deep learning approach is compatible with large-scale genomic data, paving the way for improved gene structure analysis.