PODKAT: a non-burden test for associating complex traits with rare and private variants

High-throughput sequencing technologies have facilitated the identification of large numbers of single-nucleotide variations (SNVs), many of which have already been proven to be associated with diseases or other complex traits. Since association tests considering individual SNVs independently are known to be underpowered, different collapsing strategies have be en proposed to consider multiple SNVs occurring in a region simultaneously. Such strategies can be classified into burden tests and non-burden tests, an important representative of which is the acclaimed Sequence Kernel Association Test (SKAT) by Wu et al. Several large sequencing studies, such as, the 1000 Genomes Project, the UK10K project, or the NHLBI-Exome Sequencing Project, have consistently reported a large proportion of private SNVs, that is, variants that are unique to a family or even a single individual. Non-burden tests like SKAT are typically utilizing correlations between SNVs to increase statistical power - a strategy that is not applicable to private SNVs, since singular events are generally uncorrelated. Burden tests are potentially able to deal with private SNVs, but only if the number of private SNVs occurring in a region is correlated with the trait under consideration. Moreover, burden tests have a disadvantage if deleterious and protective SNVs occur together in the same region. We propose the Position-Dependent Kernel Association Test (PODKAT). By means of a position-dependent kernel approach, PODKAT can potentially detect associations of rare and private SNVs with the trait under consideration even if the burden scores are not correlated with the trait. PODKAT assumes that, the closer two SNVs are on the genome, the more likely they have similar effects on the trait under consideration.