RPIPCM: A deep network model for predicting lncRNA-protein interaction based on sequence feature encoding.

Area of Science:

• Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• Long non-coding RNA-protein interactions (LPIs) are crucial for biological regulation.
• Predicting LPIs is challenging due to data imbalance and the need for complex feature extraction.

Purpose of the Study:

• To develop a novel deep network model, RPIPCM, for predicting LPIs.
• To address the issue of unbalanced positive and negative samples in LPI prediction.
• To evaluate the effectiveness of sequence feature encoding for LPI prediction.

Main Methods:

• Utilized a deep network model (RPIPCM) incorporating sequence feature encoding for both RNA and protein.
• Implemented a negative sampling sliding window method to handle data imbalance.
• Validated the model on diverse datasets (RPI488, ATH948, ZEA22133) and compared performance against existing methods.

Main Results:

• RPIPCM demonstrated significant improvements in accuracy, recall, precision, specificity, and MCC across multiple datasets.
• Sequence feature encoding outperformed direct original sequence encoding, showing increased predictive power.
• Comparative experiments confirmed the effectiveness and robustness of RPIPCM, especially in addressing data imbalance.

Conclusions:

• RPIPCM effectively predicts LPIs by automatically mining sequence features without external knowledge.
• The model's low cost and high efficiency make it a valuable tool for biomedical researchers.
• The proposed negative sampling method is a viable solution for data imbalance in LPI research.