Predicting functional variants in enhancer and promoter elements using RegulomeDB.

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Identifying functional genetic variants in noncoding DNA is crucial for understanding gene regulation and disease.
• Existing methods often struggle to accurately predict the impact of variants in regulatory elements like enhancers and promoters.

Purpose of the Study:

• To develop and validate a computational model, Score of Unified Regulatory Features (SURF), for predicting functional variants in enhancer and promoter regions.
• To assess the performance of SURF in predicting variant effects on gene expression and allele-specific transcription factor binding.

Main Methods:

• Trained SURF on data from massively parallel reporter assays to predict variant effects on reporter expression levels.
• Integrated features from functional genomics data (e.g., DNase footprints, ChIP-seq) and DNA sequence-based deep learning models.
• Evaluated SURF's performance in the Fifth Critical Assessment of Genome Interpretation 'Regulation Saturation' challenge.

Main Results:

• SURF achieved top performance in the 'Regulation Saturation' challenge, demonstrating high accuracy in predicting functional variants.
• Incorporating functional genomics annotations (e.g., DNase footprints, ChIP-seq data) significantly improved prediction accuracy beyond deep learning sequence features.
• The model showed good performance in predicting allele-specific transcription factor binding events.

Conclusions:

• SURF is a powerful computational tool for prioritizing functional variants in regulatory DNA.
• The integration of diverse functional genomics data enhances the predictive power of variant effect models.
• SURF can serve as an extension to existing scoring systems like RegulomeDB, advancing the understanding of noncoding variants in disease.