MIG Seminar Series - Doug Speed - Better understanding the genetic architecture of complex traits from summary statistics
Seminar/Forum

Turner Theatre
Biosciences 2
Professor's Walk
More information
T: 8344 0707
https://research.unimelb.edu.au/integrative-genomics#news-events-and-seminars
Better understanding the genetic architecture of complex traits from summary statistics
We have recently developed SumHer, software for analysing summary statistics from genome-wide association studies (GWAS). By applying SumHer to data from 24 large-scale studies (average sample size 121,000), we show that GWAS have tended to over-correct for confounding, and as a result under-reported the number of causal associations by about a quarter. We show that on average, SumHer estimates of SNP heritability are about twice those from LD Score Regression, while SumHer also gives much more modest estimates of heritability enrichments. By using the results of SumHer we can improve prediction of complex traits, and also construct a bespoke genome-wide significant threshold level, which takes into account that some SNPs are more important than others.
Presenter
-
Dr Doug Speed, Assistant Professor, Aarhus Institute of Advanced Studies, Aarhus University, Denmark
Dr Doug Speed
Assistant Professor, Aarhus Institute of Advanced Studies, Aarhus University, Denmark
Aarhus UniversityI have a strong background in mathematics and statistical genetics: my undergraduate degree was in mathematics (University of Oxford), while for my PhD I developed Bayesian methods for the analysis of genomewide association study data (University of Cambridge). From 20102017, I was at UCL, first as a postdoctoral fellow, then under a MRC Career Development Fellowship. My main focus is developing statistical methods for improved analysis of association study data. I have released the software suite LDAK, which enables more accurate estimation of SNP heritabilities, and includes MultiBLUP, the worldleading tool for constructing SNPbased prediction models. I am additionally involved in a number of worldwide analysis consortia, including for epilepsy, breast cancer and Type 1 Diabetes.