Deep genomic models of allele-specific measurements.

DeepAllele predicts allele-specific gene regulation changes with high accuracy, identifying cis-regulatory motifs across more genomic regions than existing models.

• Why it matters: Understanding how DNA variations influence gene regulation is crucial for decoding genetic contributions to traits and diseases, yet current methods rely mainly on statistical associations, limiting causal insights.
• What they did: The team developed DeepAllele, a deep learning sequence-to-function model trained on paired allele-specific data from controlled crosses and long-read sequencing, to learn sequence features affecting regulation.
• The result: DeepAllele successfully identified cis-regulatory grammar aligned with known biological mechanisms, enabling more comprehensive discovery of functional regulatory motifs in genomic studies.