Workshop in Biostatistics

DATE: January 28, 2016
TIME: 1:30 - 3:00 pm
LOCATION: Medical School Office Building, Rm x303
TITLE: Harnessing the unseen for next generation population genomics and epigenomics
SPEAKER: James Zou
Microsoft Research New England and MIT, MA      

 

Abstract:

Sequencing of large human populations has the potential to transform disease diagnosis and treatment. In order to harness the full power of this data avalanche, it is crucial to model and leverage the information and covariates that we do not see. I will illustrate this concept with examples from genomics and epigenomics. I will first discuss a close collaboration with the largest exome sequencing consortium (ExAC), where we have aggregated high-quality protein coding sequences (exomes) of 60K healthy individuals. Even with this large dataset, we can identify only a small fraction of the potentially harmful mutations that exist in the human population, because most of these variants are very rare. We developed a new algorithm that leverages the sequenced individuals to accurately infer statistical properties of the unseen genetic variation. This approach has strong mathematical guarantees and provides a unified framework to quantify the natural selection acting on our genome, annotate disease variants, and predict the discovery rate of future sequencing projects.
 
If time allows, I will describe complementary work to identify changes in the packing and chemical modifications of DNA across individuals—i.e., epigenomic variation—that are associated with diseases. This work requires flexible models of unseen covariates, especially of cell-type composition, to make valid statistical estimation.

Joint work with Greg Valiant, Paul Valiant, Daniel MacArthur and Jennifer Listgarten.

Suggested readings:

Zou, et al. (2015)  Quantifying the unobserved protein-coding variants in human populations provides a roadmap for large-scale sequencing projects.   BioRxiv, Cold Spring Harbor Laboratory.

Zout, et al. Epigenome-wide association studies without the need for cell-type composition.  Nature Methods 11, 309-311 (2014).