GERV: a statistical method for generative evaluation of regulatory variants for transcription factor binding

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Most disease-associated variants are in noncoding regulatory regions of the genome.
• Interpreting the functional consequences of these variants is crucial for genome-wide association studies (GWAS).
• Accurate variant interpretation aids in identifying causal variants linked to diseases.

Purpose of the Study:

• To introduce GERV (generative evaluation of regulatory variants), a novel computational method.
• To predict regulatory variants that impact transcription factor binding.
• To enhance the functional annotation and prioritization of causal variants.

Main Methods:

• GERV utilizes a k-mer-based generative model trained on ChIP-seq and DNase-seq data.
• It scores variants by calculating the change in predicted transcription factor binding (ChIP-seq reads) between reference and alternate alleles.
• The model captures sequence determinants, including canonical and co-factor motifs.

Main Results:

• GERV outperforms existing methods in predicting allele-specific transcription factor binding.
• It successfully identifies a validated causal variant among linked single-nucleotide polymorphisms (SNPs).
• GERV prioritizes variants known to modulate FOXA1 binding in breast cancer cell lines.

Conclusions:

• GERV offers a powerful computational approach for functional variant annotation.
• The method aids in prioritizing regulatory variants for experimental validation.
• GERV advances the interpretation of noncoding variants in genetic studies.