Accurate microRNA Target Prediction Using Detailed Binding Site Accessibility and Machine Learning on Proteomics...

Area of Science:

• Molecular Biology
• Bioinformatics

Background:

• MicroRNAs (miRNAs) regulate gene expression by targeting messenger RNA (mRNA).
• Computational miRNA target prediction methods often miss true targets and produce false positives.
• Accurate prediction of miRNA targets is crucial for understanding gene regulation.

Purpose of the Study:

• To develop a novel algorithm, DIANA-microT-ANN, for more accurate miRNA target prediction.
• To improve upon existing computational methods by integrating diverse target site features.
• To provide a reliable tool for identifying functional miRNA-mRNA interactions.

Main Methods:

• Developed DIANA-microT-ANN, an artificial neural network (ANN) algorithm.
• Incorporated novel target site features: binding structure, conservation, and accessibility.
• Trained the ANN using high-throughput experimental data on protein level changes after miRNA overexpression.
• Integrated features to generate a targeting score reflecting protein repression fold change.

Main Results:

• DIANA-microT-ANN outperforms widely used miRNA target prediction algorithms.
• The algorithm predicts a significant number of unique and reliable targets missed by other methods.
• For 542 human miRNAs, DIANA-microT-ANN identified 120,000 targets not predicted by TargetScan 5.0.
• The predicted targeting score accurately reproduces experimental protein repression levels.

Conclusions:

• DIANA-microT-ANN offers a significant advancement in miRNA target prediction accuracy and reliability.
• The algorithm enhances the discovery of functional miRNA-mRNA interactions.
• DIANA-microT-ANN provides a valuable, freely available resource for miRNA research.