TargetNet

TargetNet predicts functional microRNA (miRNA) target sites on messenger RNAs (mRNAs) to identify regulatory miRNA-mRNA interactions.

Key Features:

• Relaxed Candidate Target Site (CTS) Selection Criteria: Implements relaxed CTS selection criteria that accommodate irregularities within the seed region and extend beyond canonical site types.
• Novel Sequence Encoding Scheme: Encodes miRNA-CTS sequences using extended alignments of the seed regions to capture more comprehensive interaction information.
• Deep Residual Network-Based Prediction Model: Employs a deep residual network trained on miRNA-CTS pairs and evaluated on miRNA-mRNA pair datasets to predict functional targets and automatically extract complex features.

Scientific Applications:

• Functional miRNA target classification: Improves classification of functional miRNA targets compared with previous algorithms.
• Gene regulation and regulatory network analysis: Facilitates analysis of miRNA-mediated gene regulation and reconstruction of regulatory networks governed by miRNAs.
• Therapeutic target identification: Supports identification of high-functional targets for studying disease mechanisms and potential therapeutic strategies.

Methodology:

Training a deep residual network on curated datasets of miRNA-CTS pairs, validating on independent miRNA-mRNA pair datasets, and using a novel sequence encoding scheme together with relaxed CTS selection criteria based on extended seed-region alignments.