miRlastic

miRlastic infers miRNA-mRNA regulatory interactions from matched transcriptomic data combined with prior sequence-based target prediction resources and annotates target gene functions by exploiting the local structure of the inferred regulatory network.

Key Features:

• Inference of miRNA-Target Interactions: Uses matched transcriptomic data and prior sequence-based target prediction resources with linear regression modeling and elastic net regularization for feature selection to predict miRNA-mRNA regulatory relationships.
• Functional Annotation of Target Genes: Assigns functional roles to genes targeted by miRNAs by exploiting the local structure of the inferred regulatory network.
• Local Enrichment Analysis Procedure: Performs a local enrichment analysis that evaluates shortest paths between genes to identify regions of high functional similarity within the network.
• Performance and Validation: Validated on synthetic data with superior performance compared to commonly used methods at low sample sizes and evaluated on a head and neck squamous cell carcinoma (HNSCC) cohort from The Cancer Genome Atlas focusing on human papilloma virus (HPV)-associated miRNAs.

Scientific Applications:

• Functional annotation of miRNA targets: Assigns functions to miRNA-targeted genes based on inferred network topology and local enrichment.
• Characterization of pathway regulation: Identifies regulatory modules and pathway-level dysregulation from matched miRNA and mRNA data.
• Study of HPV-associated dysregulation in HNSCC: Infers miRNA-mRNA regulatory networks and functional miRNA clusters enriched for experimentally validated interactions in TCGA HNSCC cohorts focusing on HPV-associated miRNAs.

Methodology:

Fits linear regression models with elastic net regularization using transcriptomic data and prior sequence-based target predictions, infers an miRNA-mRNA regulatory network, and applies a local enrichment analysis evaluating shortest paths between genes to assign functional annotations.