Predicting the interaction biomolecule types for lncRNA: an ensemble deep learning approach.

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Long noncoding RNAs (lncRNAs) are crucial regulators in biological processes and disease.
• Current in silico methods for lncRNA function prediction often rely on similarity or specific interaction analysis.
• A need exists for methods that predict the type of biomolecule a lncRNA interacts with.

Purpose of the Study:

• To develop a computational tool for predicting the interaction biomolecule type of a given lncRNA.
• To investigate the molecular mechanisms underlying lncRNA interactions with DNA, RNA, and proteins.
• To provide a novel perspective on understanding lncRNA functions through interaction prediction.

Main Methods:

• Investigated molecular mechanisms of lncRNA-RNA, lncRNA-protein, and lncRNA-DNA interactions.
• Developed an ensemble deep learning model named lncIBTP (lncRNA Interaction Biomolecule Type Prediction).
• Evaluated model performance using 5-fold cross-validation.

Main Results:

• The lncIBTP model achieved an average accuracy of 0.7042.
• Macro-average Area Under the ROC Curve (AUC) was 0.7903, and Area Under the Precision-Recall Curve (AUPRC) was 0.6421.
• Analysis suggested potential distinct characteristics for lncRNAs interacting with DNA versus RNA.

Conclusions:

• The lncIBTP model effectively predicts the type of biomolecule a lncRNA interacts with.
• The findings highlight the potential for differentiating lncRNA interaction mechanisms.
• This work offers a new approach to exploring lncRNA functions and their roles in biological systems.