miRBind

miRBind predicts miRNA:target-site binding using a deep learning classifier trained on seed-agnostic experimental data to identify canonical and non-canonical interactions mediated by the Argonaute (Ago) protein family.

Key Features:

• Deep learning model: Uses a deep learning classifier to perform miRNA:target-site binding classification.
• Seed-agnostic training: Trained on seed-agnostic experimental data to capture interactions that do not follow canonical seed rules.
• Canonical and non-canonical detection: Identifies both canonical seed-based and non-canonical binding sites.
• Argonaute-mediated interactions: Models interactions that are mediated by the Argonaute (Ago) protein family.
• Comparison to traditional approaches: Moves beyond seed-based heuristics and co-folding free-energy calculations and reports improved prediction accuracy over those methods.
• Binding classification focus: Produces binary or probabilistic predictions of miRNA:target-site binding potential.

Scientific Applications:

• miRNA-mediated gene regulation: Improves identification of miRNA targets for studies of post-transcriptional gene regulation.
• Disease mechanism investigation: Facilitates analysis of miRNA involvement in disease-related regulatory changes.
• Therapeutic target discovery: Supports discovery and prioritization of miRNA-related therapeutic targets.
• Experimental prioritization: Prioritizes candidate binding sites, including non-canonical sites, for experimental validation.

Methodology:

Trains a deep learning classifier on seed-agnostic experimental data to classify miRNA:target-site binding while not relying on seed-centric heuristics or co-folding free-energy calculations.