DM-RPIs: Predicting ncRNA-protein interactions using stacked ensembling strategy

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• ncRNA-protein interactions (ncRPIs) are crucial for cellular functions including gene regulation and disease development.
• Experimental identification of ncRPIs is often resource-intensive and time-consuming.
• Developing efficient computational methods is essential for advancing the study of ncRPIs.

Purpose of the Study:

• To propose a novel computational method, Deep Mining ncRNA-Protein Interactions (DM-RPIs), for predicting ncRNA-protein interactions.
• To leverage deep learning and ensemble strategies for enhanced prediction accuracy.
• To provide a cost-effective and time-efficient alternative to experimental ncRPI identification.

Main Methods:

• Utilized Deep Stacking Auto-encoders Networks (DSANs) to reduce dimensionality and extract features from RNA and protein sequences based on k-mer frequency.
• Trained individual predictors using Support Vector Machine (SVM), Random Forest (RF), and Convolution Neural Network (CNN).
• Integrated the individual predictors using a stacked ensembling strategy for a final prediction model.

Main Results:

• The DM-RPIs method achieved a high accuracy of 0.851 on the RPI2241 dataset.
• Key performance metrics included precision (0.852), sensitivity (0.873), specificity (0.826), and Matthews Correlation Coefficient (MCC) of 0.701.
• The results demonstrate the efficacy and potential of the proposed computational approach.

Conclusions:

• DM-RPIs presents a promising and pioneering computational tool for predicting ncRNA-protein interactions.
• The study highlights the effectiveness of combining deep learning feature extraction with stacked ensemble methods.
• This method offers a valuable resource for researchers studying ncRPIs and their roles in biological processes and diseases.