Taking the guesswork out of genetic analysis

Taking the guesswork out of genetic analysis

22 Sep 2014 - 14:33

Genome-wide association studies (GWAS) and other experiments can now provide more accurate information about the heritability and genetic causes of traits such as disease risk. WarpedLMM, a new powerful method developed by EMBL-EBI and Microsoft Research, automates a previously manual process, improving results and removing an element of bias from genetic analyses. The method is published in Nature Communications.

Researchers who study the genetics of disease often work with very large and tangled datasets that are full of confounding variables. To make sense of these data, statistical tools called Linear Mixed Models (LMMs) are used to identify which genetic differences between people are relevant to disease, and to estimate the strength of their effects. While LMMs are robust, their application in practice can prove difficult. For example, researchers must pre-process their phenotype data, and that means quantifying (manually) all of the phenotypic features in the samples they would like to study. Because it is not immediately clear exactly how to do this in the best way, the process is not objective and results in reduced statistical power and biased estimates.

The new method, WarpedLMM, provides a solution to this problem by automatically processing different phenotypes, such that the models work most effectively, removing the element of guess-work by individual researchers.

“We’ve developed new statistical methods that are really useful for genome wide association studies and heritability estimation, which investigate the genetic underpinning of important phenotypes such as human diseases,” explains Oliver Stegle, Research Group Leader at EMBL-EBI.

Nicolo Fusi of Microsoft Research adds, “WarpedLMM helps researchers attribute a greater proportion of phenotypic differences between individuals to genetics – connections they might otherwise have missed. It is a practical improvement to a method that is already widely used, and it can help create a more accurate picture of the genetics of quantitative traits.”

The researchers tested their model using simulations, followed by analysis of datasets from both humans and model organisms. They showed that their model increases power in genome-wide association studies, improves the accuracy of heritability estimation and allows for more accurate prediction of phenotypes.

Source article: Fusi, N., Lippert, C., Lawrence, N.D. and Stegle, O. (2014) Warped linear mixed models for the genetic analysis of transformed phenotypes. Nature Comm (in press). Published online 19 September 2014; doi: 10.1038/ncomms5890

Software: An implementation of WarpedLMM is available at http://github.com/pmbio/warpedLMM

Funding: The NFBC1966 Study is conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with the Broad Institute, UCLA, University of Oulu, and the National Institute for Health and Welfare in Finland. O.S. was supported by a Marie Curie FP7 fellowship.

Contact the news team

Oana Stroe
Senior Communications Officer
+44 (0)1223 494 369

Subscribe to the email newsletter

Subscribe to our publications.

Sign up Or stay updated with the RSS feed (EMBL-EBI only).