Dr Doug Speed

University College London

 FRIDAY, 4^TH JULY, 2014

12.00 – 13.00

Room OS6 – Oakfield House

"SNP-based heritability analysis - is there anything it can't do?"

 

Abstract

The primary use of SNP-based heritability analysis is to calculate how much of a trait's phenotypic variation is explained by common SNPs. For example, the first application showed that common variants explain the majority of heritability for human height, with similar results found for intelligence, schizophrenia, Parkinson's, BMI, eating disorders, cannabis use, lifespan, to name just a few. However, calculating variance explained is just the tip of iceberg. In this talk, I will start by outlining half a dozen ways SNP-based heritability analysis can be used to investigate the genetic architecture of complex traits; from providing a gene-based test of association, to measuring intensity of heritability for functional categories of SNPs; from testing concordance between traits to classifying heterogeneous diseases. Then I will describe MultiBLUP, a novel tool for SNP-based prediction. MultiBLUP retains the computational advantages of BLUP but provides a huge improvement in prediction accuracy. I find that for each of nine human diseases, and across 139 mice phenotypes, MultiBLUP outperforms all rival methods considered, and computationally is over 10 times more efficient than the next best-performing method.

 Doug Speed

Doug was an undergraduate in maths before undertaking a PhD with Simon Tavare in Cambridge, where he worked on methodology for identifying interactions in genome-wide association studies. In 2010 he moved to University College London to work with David Balding, where his main area is developing tools for SNP-based heritability analysis with focus on applying these to epilepsy. In January he will continue this work as a MRC Early Career Research Fellow in Biostatistics.

  

 

ALL WELCOME