ncDENSE: a novel computational method based on a deep learning framework for non-coding RNAs family prediction.

Area of Science:

• Life Sciences
• Bioinformatics
• Computational Biology

Background:

• Non-coding RNAs (ncRNAs) are crucial in biological processes, but their functions are often unknown.
• Accurate ncRNA family prediction is essential for elucidating ncRNA functions, as related ncRNAs share similar roles.
• Current prediction methods, based on secondary structures or sequence features, face challenges with complexity and accuracy.

Purpose of the Study:

• To develop a novel, highly accurate method for predicting ncRNA families.
• To leverage ncRNA sequence features for improved family classification.
• To address the limitations of existing ncRNA family prediction techniques.

Main Methods:

• Proposed ncDENSE, a deep learning model utilizing ncRNA sequence features.
• Employed one-hot coding to encode ncRNA sequences.
• Integrated dynamic Bi-directional Gated Recurrent Unit (Bi-GRU), Dense Convolutional Network (DenseNet), and Attention Mechanism (AM) within an ensemble model.

Main Results:

• ncDENSE demonstrated superior performance in ncRNA family prediction.
• The model achieved significant improvements in Accuracy, Sensitivity, Precision, F-score, and Matthews Correlation Coefficient (MCC).
• Specific performance gains ranged from 2.08% to 2.39% compared to suboptimal methods.

Conclusions:

• The ncDENSE method effectively predicts ncRNA families by extracting sequence features.
• The integration of dynamic Bi-GRU, DenseNet, and AM enhances prediction accuracy.
• This study provides a more accurate tool for ncRNA family identification and functional prediction.