Incorporating between-pedigree co-ancestry in variance-components linkage
analysis
James E. Hicks, Michael A. Province
As rare variants become an increasing focus in genetic epidemiology, family-based
study designs will return to the forefront. Tools, such as linkage analysis, used for
family data will need to be updated to accommodate this hypothesis.
In conventional linkage analysis pedigrees are assumed to be independent, and only
identity-by-descent (IBD) states within each pedigree are used. However, cryptic
relatedness is present in populations and haplotypes shared by cryptically related
individuals can harbor rare variants that influence phenotypes.
With the development of long range phasing-based methods for detection of shared
genomic segments using dense genotypes, IBD states across the genome can be
inferred, without use of pedigree information. This is done by identifying long runs
of genotypes identical-by-state which are unlikely to be identical without being IBD.
This method allows for cryptic relatedness to be incorporated into linkage analysis.
In variance-components linkage analysis, IBD states are modeled in a covariance
matrix. Conventionally, these states are computed within pedigrees by modeling
recombination rates between genotypes in a sparse set of markers. Replacing that
covariance matrix with one determined from shared segment methods can increase
the accuracy and power of linkage analysis. Power is increased in the scenario when
there is a haplotype shared IBD between members of different pedigrees. If there is
no between-pedigree IBD, the analysis reduces to conventional variancecomponents analysis. By determining IBD states by long runs of dense IBS
genotypes, linkage signals can be determined from their physical position, allowing
more precise localization.