CharID

CharID predicts chromatin loops mediated by open chromatin regions (OCRs) to identify OCR anchors and their interactions for studying gene expression regulation.

Key Features:

• Two-Step Predictive Model: CharID implements a two-step predictive model in which CharID-Anchor uses an attention-based hybrid Convolutional Neural Network (CNN) and Bidirectional Gated Recurrent Unit (BiGRU) to distinguish anchor versus non-anchor OCRs, and CharID-Loop uses a gradient boosting decision tree with a chromosome-split strategy to predict interactions between identified anchor OCRs.
• Anchor-Type Versatility: Detects a wide variety of chromatin loop types without restriction to specific anchor types such as enhancer–promoter or architectural protein-mediated loops.
• Performance Benchmarking: Demonstrated superior predictive performance compared to other algorithms across three human cell lines.
• OCR-Mediated Interaction Networks: Enables construction of OCR-mediated interaction networks and identification of hub anchors often located near housekeeping genes.
• SNP-Loop Analysis: Supports analysis of loops containing disease-associated single nucleotide polymorphisms (SNPs), including identification of a SNP–gene loop implicating GFOD1 in cardiovascular disease.

Scientific Applications:

• Chromatin Architecture Mapping: Mapping OCR-mediated chromatin architecture to investigate regulatory interactions that influence gene expression.
• Network Biology: Constructing interaction networks to identify hub anchors and link chromatin loops to housekeeping gene regulation.
• Variant Interpretation: Interpreting noncoding disease-associated SNPs by connecting variants to putative target genes via predicted loops (e.g., GFOD1 in cardiovascular disease).

Methodology:

CharID applies a two-step computational framework: CharID-Anchor uses an attention-based hybrid CNN and BiGRU to classify anchor OCRs, and CharID-Loop uses a gradient boosting decision tree with a chromosome-split strategy to predict interactions between anchors.