An explainable hybrid CNN-LSTM framework for accurate sequence-based classification of RNA N6-methyladenosine (m6A)...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• N6-methyladenosine (m6A) is a crucial RNA modification impacting gene regulation.
• Identifying m6A sites is challenging due to sequence complexity and data limitations.

Purpose of the Study:

• To develop an accurate and interpretable method for identifying RNA m6A modification sites.
• To enhance post-transcriptional gene regulation analysis through improved m6A site prediction.

Main Methods:

• A hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) framework was developed.
• Shapley Additive Explanations (SHAP) was integrated for feature selection and model interpretability.
• RNA sequences were encoded using biologically relevant features, processed by CNN for spatial features and LSTM for temporal dependencies.

Main Results:

• The CNN-LSTM model with SHAP feature selection outperformed traditional and standalone deep learning models.
• Achieved high performance metrics: 87.39% accuracy, 83.25% sensitivity, 91.52% specificity, and 0.7534 MCC.
• Demonstrated strong potential for transcriptome-wide epitranscriptomics analysis.

Conclusions:

• The proposed CNN-LSTM-SHAP framework accurately classifies RNA m6A sites.
• The model offers enhanced biological insight and predictive performance for RNA modifications.
• This approach advances the field of epitranscriptomics research and analysis.