Predicting the impact of rare regulatory variation

The enormous increase in availability of full human genomic sequences presents great opportunity for understanding the impact of rare genetic variants. Based on current knowledge, however, we are still limited in our ability to interpret or predict regulatory consequences of rare variants. Association methods are inherently limited for variants observed in only one or a few individuals. Diverse genomic annotations such as growing resources of epigenetic data and have been shown to be informative regarding regulatory elements, but are only moderately predictive of impact for individual variants. The availability of personal RNA-seq and other cellular measurements for the same individuals with genome sequencing offers a new avenue for integrated methods for prioritizing rare functional variants. By considering informative genomic annotations along with molecular phenotyping, we are able to identify the regulatory impact of rare genetic variants largely excluded from previous analyses. We have evaluated the impact of rare regulatory variants using whole genome sequences and corresponding RNA-sequence in 44 different tissues samples from the GTEx project and report.  We demonstrate that rare variants in promoter elements, conserved regions, and variants that affect splicing are associated with extreme changes in gene expression across multiple tissues. Additionally, we have developed a Bayesian machine learning approach that integrates whole genome sequencing with RNA-seq data from the same individual, leveraging gene expression levels along with diverse genomic annotations and performing joint inference to identify likely functional rare regulatory variants. We have applied this model to the GTEx data to prioritize the most likely functional rare regulatory variants for each individual. We demonstrate that integrative models perform better than predictions from DNA-sequencing or RNA-sequencing alone. Our probabilistic model of rare regulatory genetic variants offers potential for identifying deleterious regulatory variants from individual genomes.

 

Alexis Battle’s research focuses on unraveling the impact of genetics on the human health, using machine learning and probabilistic methods to analyze large scale genomic data.  She is interested in applications to personal genomics, genetics of gene expression, and gene networks in disease, leveraging diverse data to infer more comprehensive models of genetic effects on the cell.  She earned her Ph.D. and Masters in Computer Science in 2014 from Stanford University in 2014, where she also received her Bachelors in Symbolic Systems (2003).  Alexis also spent several years in industry as a member of the technical staff at Google.  Prior to joining Hopkins, Alexis spent a year as a postdoc with Jonathan Pritchard with HHMI and the Genetics Department at Stanford.  She joined JHU in July 2014.