A heterogeneous information network learning model with neighborhood-level structural representation for predicting...

Area of Science:

• Biotechnology and Bioinformatics
• Genomics and Molecular Biology

Background:

• Long non-coding RNAs (lncRNAs) and microRNAs (miRNAs) are crucial in human disease mechanisms.
• Experimental identification of lncRNA-miRNA interactions (LMIs) is laborious and time-consuming.
• Existing computational methods often overlook complex biomolecular mechanisms, limiting prediction accuracy.

Purpose of the Study:

• To develop an advanced computational model for precise prediction of lncRNA-miRNA interactions (LMIs).
• To leverage heterogeneous information networks (HINs) for improved LMI prediction accuracy.
• To integrate biological knowledge and network topology for a comprehensive understanding of LMIs.

Main Methods:

• Construction of a heterogeneous information network (HIN) integrating nine types of biomolecular interactions.
• Application of representation learning strategies to derive biological and network embeddings for lncRNAs and miRNAs.
• Utilization of the XGBoost classifier with learned embeddings for predicting unknown LMIs.

Main Results:

• HINLMI demonstrated superior performance compared to state-of-the-art models on real-world datasets.
• The model accurately predicted LMIs by simultaneously considering biological knowledge and network topology.
• Analysis confirmed the effectiveness of integrating rich heterogeneous information for identifying novel LMIs.

Conclusions:

• HINLMI offers a powerful and accurate computational approach for predicting lncRNA-miRNA interactions.
• Integrating diverse biomolecular data within a HIN framework enhances LMI prediction.
• This method provides valuable insights for discovering novel lncRNA-miRNA interactions relevant to human diseases.