A Multi-Input Neural Network Model for Accurate MicroRNA Target Site Detection.

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• MicroRNAs (non-coding RNA sequences) regulate gene expression by targeting messenger RNAs.
• Identifying microRNA target sites is crucial for understanding cellular functions but is challenging due to experimental costs and computational limitations.
• Existing computational methods often suffer from high false-positive rates.

Purpose of the Study:

• To develop a novel computational algorithm for accurate microRNA target site prediction.
• To overcome the limitations of current prediction methods by integrating diverse biological features.

Main Methods:

• Introduction of a Multi-Input Neural Network (MINN) algorithm.
• Integration of biologically relevant features: microRNA duplex structure, substructures, minimum free energy, and base-pairing probabilities.
• Representation of features as images processed in parallel by the MINN for comprehensive learning.

Main Results:

• The MINN algorithm achieved high performance on an experimentally validated test set.
• Key metrics include an Area Under the Precision-Recall Curve (AUPRC) of 0.9373, Precision of 0.8725, and Recall of 0.8703.
• The proposed method outperformed several commonly used computational microRNA target-site prediction tools.

Conclusions:

• Integrating diverse, biologically interpretable features (duplex structure, MFE, binding probabilities) enhances prediction accuracy.
• The model demonstrates strong generalization capabilities, performing well on sequentially distant samples.
• Focusing on structural and energetic features, rather than solely nucleotide sequences, improves model robustness.