Genome-scale MicroRNA target prediction through clustering with Dirichlet process mixture model

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• MicroRNA (miRNA) regulation fine-tunes human gene networks and is involved in numerous physiological and pathological conditions.
• Current computational tools for miRNA-mRNA interaction prediction primarily rely on sequence pairing, limiting their practical application due to complex binding dynamics.
• The interplay of competing and cooperative miRNA binding significantly complicates regulatory processes, hindering accurate target prediction.

Purpose of the Study:

• To develop an improved computational method for microRNA target prediction.
• To address the limitations of existing tools by incorporating diverse molecular features and accounting for complex binding interactions.
• To enhance the accuracy and reduce false positives in predicting miRNA binding sites, particularly for transcript isoforms.

Main Methods:

• Developed a novel prediction model based on the Dirichlet Process Gaussian Mixture Model (DPGMM).
• Integrated a comprehensive set of molecular features related to miRNA, mRNA, and their interaction sites.
• Validated the model using large-scale sequencing data and screened the entire human transcriptome.

Main Results:

• The DPGMM-based model demonstrated superior predictive performance compared to several state-of-the-art tools.
• Achieved higher accuracy in identifying transcript isoform-specific binding sites with a significant reduction in false positives.
• Successfully applied the predicted targets to construct conditional miRNA-mediated gene regulatory networks in human cancer.

Conclusions:

• The probability-based prediction method offers a valuable tool for distinguishing miRNA targets based on binding potential.
• Provides enhanced capability for investigating dynamic gene regulation, especially in scenarios involving binding competition.
• Facilitates a deeper understanding of miRNA's role in complex biological processes and disease states.