MetaChrom

MetaChrom predicts the regulatory effects of non-coding genetic variants on epigenomic profiles at single-nucleotide resolution using deep transfer learning.

Key Features:

• Deep Transfer Learning Framework: integrates extensive reference datasets of chromatin and epigenomic data with phenotype-specific epigenomic profiles, including profiles relevant to neuropsychiatric disorders.
• High-Resolution Predictions: predicts the impact of genomic variants on epigenomic features at single-nucleotide resolution.
• Phenotype-Specific Accuracy: trains and evaluates models using datasets from fetal and adult brain tissues and cellular models representing early neurodevelopment to improve accuracy in phenotype-relevant contexts.
• Functional Variant Prediction: identifies functional variants and is validated against experimentally determined regulatory variants from induced pluripotent stem (iPS) cell-derived neurons.
• Integration with GWAS Data: facilitates integration of genome-wide association study (GWAS) data to prioritize candidate single nucleotide polymorphisms (SNPs) for diseases such as Schizophrenia (SCZ).

Scientific Applications:

• Gene regulation: study mechanisms by which non-coding DNA sequences and variants modulate epigenomic features and regulatory functions.
• Neuropsychiatric disorder genetics: prioritize and interpret non-coding variants and candidate risk genes in disorders such as Schizophrenia (SCZ) using fetal/adult brain and neurodevelopmental cellular models.
• Variant interpretation and prioritization: prioritize candidate SNPs from GWAS for functional follow-up based on predicted epigenomic impact.

Methodology:

Trains and evaluates deep learning models via deep transfer learning on extensive reference chromatin and epigenomic datasets combined with phenotype-specific epigenomic profiles, using fetal and adult brain tissue datasets and cellular models of early neurodevelopment, and integrates GWAS data to prioritize SNPs.