PlncRNA-HDeep: plant long noncoding RNA prediction using hybrid deep learning based on two encoding styles.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long noncoding RNAs (lncRNAs) are crucial regulators of biological activities, necessitating accurate prediction methods.
• Existing lncRNA predictors are limited for plant species, highlighting the need for specialized tools.
• Deep learning models like Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN) can extract features from RNA sequences, but ensemble approaches offer enhanced performance.

Purpose of the Study:

• To develop a powerful and reliable predictor for plant long noncoding RNAs (lncRNAs).
• To leverage deep learning, specifically hybridizing LSTM and CNN models, for enhanced lncRNA prediction.
• To explore novel encoding strategies for RNA sequences to improve predictive accuracy.

Main Methods:

• Proposed PlncRNA-HDeep, a hybrid deep learning model integrating Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN).
• Utilized p-nucleotide and one-hot encodings for RNA sequences, treating them as sentences and images respectively.
• Optimized model performance through parameter adjustment and testing three hybrid strategies.

Main Results:

• PlncRNA-HDeep achieved high performance on the Zea mays dataset, with 97.9% sensitivity, 95.1% precision, 96.5% accuracy, and 96.5% F1 score.
• The hybrid model outperformed individual LSTM and CNN models, as well as several shallow machine learning methods and existing lncRNA prediction tools.
• Demonstrated the effectiveness of combining different encoding styles and deep learning architectures for plant lncRNA prediction.

Conclusions:

• PlncRNA-HDeep is a feasible and effective tool for predicting plant long noncoding RNAs.
• The model provides credible predictive results and demonstrates the potential of deep learning in this field.
• This approach may offer valuable insights and references for future research in lncRNA prediction.