David Geffen School of Medicine at UCLA
Department of Human Genetics

Speaker Series - Fall Quarter 2013

Mondays, 11am - 12pm, Gonda Building First Floor Conference Room, 1357

Mon, Oct 14
Human Disorders in Glycosylation: The Bountiful Harvest of Exome Analysis
Hudson Freeze, PhD, Director,Genetic Disease Program, Sanford-Burnham Medical Research Institute
Contact & Intro: Kate Neckermann, kneckermann@mednet.ucla.edu
Mon, Oct 28
Genetics-Based Targeted Therapy in Acute Myeloid Leukemia
Lucio H Castilla, Ph.D., Associate Professor, Molecular Medicine, University of Massachusetts
Contact & Intro: Esteban Dell'Angelica, edellangelica@mednet.ucla.edu
Mon, Nov 25
Using large-scale genetic data to learn about human history and variation
Joseph Pickrell, Ph.D., Harvard Medical School, Boston, Massachusetts
Contact & Intro: Kate Neckermann, kneckermann@mednet.ucla.edu
Mon, Dec 09
Powerful Bayesian mixed model association in time O(MN): is it possible?
Alkes Price, Ph.D., Assistant Professor of Statistical Genetics, Department of Epidemiology, Department of Biostatistics, Harvard School of Public Health
Contact & Intro: Kate Neckermann, kneckermann@mednet.ucla.edu
View details »

ABSTRACT: Mixed model association is an appealing approach for identifying genetic associations while avoiding confounding. Recent work has reduced the computational cost of this approach to O(MN^2) (where M = #markers and N = #samples), which is the cost of computing the underlying genetic relationship matrix (GRM). Here, we describe a method that obtains the same results in O(MN) time, without explicitly computing the GRM. Our association statistic is based on the phenotypic residual of the BLUP prediction and is calibrated via LDscore regression, which relies on the linear relationship between a marker's expected association statistic and its summed LD with other markers. We describe Bayesian extensions to the method that enable an increase in power under non-infinitesimal genetic architecture, again in O(MN) time.

Previous Quarters