GuidingNet

GuidingNet predicts genome-wide binding sites of DNA methyltransferases (DNMTs) and identifies transcriptional cofactors to study locus-specific DNA methylation and transcriptional regulation.

Key Features:

• Network-regularized logistic regression: Uses a network-regularized logistic regression model to infer binding sites from integrated data.
• Multi-omics integration: Integrates gene expression profiles, chromatin accessibility information, DNA sequences, and protein-protein interaction networks.
• TF network guidance: Models interactions between DNMTs and transcription factors to capture TF-guided DNMT binding.
• DNMT coverage: Targets DNA methyltransferases including DNMT3A, DNMT3B, and DNMT3L.
• Cross-tissue and cross-species prediction: Applied across diverse tissue types in human and mouse models.
• Performance advantage: Demonstrates improved prediction accuracy compared to single-data-source methods and approaches using sparsity regularization.
• Extensible to other chromatin regulators: Applicable to inference of binding for other chromatin regulators beyond DNMTs.
• Experimental validation: Validated against experimental methylation data for robustness within and across tissues.

Scientific Applications:

• Genome-wide DNMT binding prediction: Predicts locus-specific DNMT binding sites across the genome.
• Identification of transcriptional cofactors: Identifies transcription factors and cofactors that guide DNMT binding and methylation patterns.
• Study of DNA methylation regulation: Enables investigation of mechanisms underlying locus-specific DNA methylation and its role in gene regulation and development.
• Chromatin regulator analysis: Facilitates analysis of binding and function for other chromatin regulators.
• Cross-tissue and cross-species analysis: Supports comparative analyses of DNMT binding across tissues and between human and mouse.
• Benchmarking and validation: Serves as a framework for benchmarking predictions against experimental methylation datasets.

Methodology:

Implements a network-regularized logistic regression that integrates gene expression, chromatin accessibility, DNA sequence, and protein-protein interaction networks, models TF–DNMT interactions via the TF network, and compares performance to single-data-source and sparsity-regularized approaches.