About Eagle

Authors

Developed by Andrew W. George (Data61, CSIRO) with lots of help at critical moments from Joshua Bowden (IM&T, CSIRO) and Ari Verbyla (Data61, CSIRO).

Purpose:

Association mapping is a difficult problem. We have far more predictors/features/variables (i.e. snps) than samples. Also, the predictors are highly correlated. To simplify the problem, it is common to measure the strength of association between a snp and trait for each snp separately. This approach is also attractive, computationally. However, the Type 1 error rate (the chance of a false positive) can be inflated due to issues from multiple testing. Also, single-locus models do not reflect reality. More advanced multiple-locus models have been developed but they often have a high computational cost and their results can be difficult to interpret.

Our goal was to develop a package that implements a state-of-the-art multiple-locus method for association mapping that allows GWAS data to be analysed routinely. To this end, we have created the Eagle package. Eagle is built on the R language but you do not need to know R to use it. Functions within our package can be accessed through a web-based GUI system. Our package is easy-to-use, it can handle the analysis of data larger than the memory capacity of a computer, the GWAS can have arbitrary structure, and it harnesses multiple threads for computation. Eagle is best suited for the analysis of quantitative (continuous) trait data.