Bioinformatics Seminar
Time:
Venue: Na
15 August 2017
NaUniversity of Melbourne
John L. HopperCentre for Epidemiology and Biostatistics
Help find the "missing heritability" of breast cancer;Breast cancer became a major disease of Western women last century ; and is increasing rapidly in the developing world. Over the last decade ; a global initiative in genome wide association studies (GWAS) has used highly conservative statistical approaches to identify several hundred loci at which there are common genetic variants (SNPs) associated with risk. Despite the enormous cost and multiple high-profile papers published ; there has yet to be any substantive translational impact of these GWAS findings.
These SNPs explain only a small proportion of familial aggregation of the more common hormone receptor positive from of the disease ; and a much smaller proportion of the hormone receptor negative disease (diagnosed at a younger age ; and implicated in greater morbidity and mortality) despite it being much more strongly familial. In terms of the ability to discriminate cases from controls on a population basis (by considering the change in odds per adjusted standard deviation; Hopper ; 2015) ; the GWAS-based risk scores come behind our new semi-automated and automated measures of mammographic density and ; at least for breast cancer diagnosed before age 50 years ; risk scores based on mutigenerational family history.
So what has gone wrong ; and what can be done to address the "missing heritability" whatever that means?
We have conducted the discovery phase of a GWAS of >3 ;500 breast cancer cases enriched for likely genetic risk through having a family history and or young age at diagnosis ; and >3 ;500 older controls. More than 4 million SNPs have been measured and imputed to >50 million markers.
The Challenge is to use these data to create better risk predictors ; the ability of which will be determined by application to data from the OncoArray GWAS of more than 100 ;000 breast cancer cases and 100 ;000 controls.