MirMark

MirMark predicts microRNA (miRNA) target sites and untranslated regions (UTRs) using machine learning to improve the accuracy of miRNA-target identification for studying gene regulation and disease mechanisms.

Key Features:

• Site-Level and UTR-Level Prediction: Provides predictions at both specific binding site (site-level) and untranslated region (UTR-level) resolution.
• Machine Learning Integration: Integrates experimentally verified miRNA targets from miRecords and mirTarBase to train predictive models.
• Extensive Feature Set: Utilizes over 700 features for model training and prediction.
• Feature Selection: Employs Correlation-based Feature Selection (CFS) to identify relevant predictive features.
• Classifier Construction: Builds site- and UTR-level classifiers using a variety of statistical and machine learning methods.
• Improved Predictive Performance: Demonstrates superior accuracy compared to other publicly available miRNA target prediction methods.

Scientific Applications:

• Cancer and Disease Research: Supports identification of miRNA targets implicated in cancers and other diseases.
• Pathway and Regulatory Network Analysis: Facilitates elucidation of molecular pathways and gene regulatory networks via more accurate target prediction.
• Therapeutic Target Identification: Aids identification of potential therapeutic targets by improving confidence in predicted miRNA-target interactions.

Methodology:

MirMark trains machine learning models using experimentally verified targets from miRecords and mirTarBase, leverages a feature set of over 700 attributes, applies Correlation-based Feature Selection (CFS) to select relevant features, and uses a variety of statistical and machine learning methods to construct site- and UTR-level classifiers.