Researchers from the University of Michigan's School of Public Health have developed an approach for associating rare genetic variants with diseases called Rare Variant Analysis and Meta-Analysis (RAREMETAL) that makes use of statistical information from multiple individual-level studies.
The RAREMETAL framework, as explained in a recent Nature Genetics paper, lets users aggregate test statistics from individual rare variant studies and then analyze them using various gene-level association tests such as weighted and unweighted burden tests and the sequence kernel association test.
The benefit of RAREMETAL, according to Dajiang Liu, a postdoctoral research fellow at U of M and the first author of the paper, is that it provides a way to combine information across studies enabling users to obtain sample sizes that are large enough to identify potential associations between atypical variants and complex traits with sufficient power. Rare variants, as the name implies, occur less frequently in the population, so in order to accurately study their impacts scientists need to be able to group those that are similar and evaluate their combined effects.
Compared to alternative statistical approaches such as Fisher’s method that have been used to try to study these variants, RAREMETAL provides "great flexibility in the choice of rare variant association test;" includes "rich information that helps in interpretation;" and "allows the relationship between multiple association signals in a region to be dissected through conditional analysis," the researchers wrote. Its results, they said, are "comparable" to sharing individual-level data and "identical" when allowances are made "for between-study hetero¬geneity in nuisance parameters, such as trait means, variances, and covariate effects."
To demonstrate its efficacy, the researchers used RAREMETAL to analyze the blood lipid levels of more than 18,600 individuals drawn from seven different studies. According to the Nature Genetics paper, they were able to use their method to confirm the results reported by the individual studies, in terms of identifying rare variants, and also identified additional loci that are associated with lipid levels in humans.
The researchers believe that their method will speed up rare variant research and result in more leads for drug development efforts. For their next steps, Liu told BioInform that the team is applying RAREMETAL to a number of large-scale studies to, for example, explore glycemic traits such as body mass index and height; working on improvements to the tool based on user feedback; and trying to make the software scalable enough to handle data from larger studies.
The researchers have packaged the method in freely available software that includes tools for generating and annotating summary statistics; performing meta-analysis; calculating gene-level statistics; and executing conditional analyses.