DeepLUCIA

DeepLUCIA predicts chromatin loops from genomic and epigenomic profiles to identify tissue-specific regulatory interactions and link GWAS SNPs to putative target genes without requiring CTCF signal information.

Key Features:

• Deep learning framework: Employs a deep learning architecture to predict chromatin loops directly from genomic and epigenomic profiles without using CTCF signal or TF binding profiles.
• Tissue-specific predictions: Generates tissue-specific chromatin loop predictions by leveraging tissue-specific epigenomes.
• Independence from TF binding data: Operates without transcription factor binding profile inputs while capturing chromatin interaction signals.
• Comparable accuracy: Achieves prediction accuracies comparable to TF binding-dependent models despite not using TF binding data.
• GWAS SNP target prediction: Predicts novel target genes for single nucleotide polymorphisms identified in genome-wide association studies.
• Disease applications: Has been applied to SNPs associated with Brugada syndrome, COVID-19 severity, and age-related macular degeneration.

Scientific Applications:

• Tissue-specific gene regulation: Mapping chromatin loops to study regulatory mechanisms of gene expression across human tissues.
• GWAS interpretation: Linking disease-associated SNPs to candidate target genes through predicted chromatin interactions.
• Disease mechanism investigation: Prioritizing target genes and regulatory interactions in conditions such as Brugada syndrome, COVID-19 severity, and age-related macular degeneration.

Methodology:

Uses a deep learning architecture trained on genomic and epigenomic profiles to predict chromatin loops directly without requiring CTCF signal or TF binding profiles.