Predicting RNA-seq coverage from DNA sequence as a unifying model of gene regulation

Sequence-based machine-learning models trained on genomics data improve genetic variant interpretation by providing functional predictions describing their impact on the cis-regulatory code. However, current tools do not predict RNA-seq expression profiles because of modeling challenges. Here, we introduce Borzoi, a model that learns to predict cell-type-specific and tissue-specific RNA-seq coverage from DNA sequence. Using statistics derived from Borzoi's predicted coverage, we isolate and accurately score DNA variant effects across multiple layers of regulation, including transcription, splicing and polyadenylation. Evaluated on quantitative trait loci, Borzoi is competitive with and often outperforms…

• NI
  National Institutes of Health
• NH
  National Heart, Lung, and Blood Institute
• NI
  National Institute of Mental Health
• NI
  National Institute on Drug Abuse
• NH
  National Human Genome Research Institute
• NO
  NIH Office of the Director
• NC
  National Cancer Institute
• NI
  National Institute of Neurological Disorders and Stroke
• CF
  Common Fund