Computer-Simulation Methods in Human Linkage Analysis
J. Ott
Proceedings of the National Academy of Sciences of the
United States of America
86(11), 4175--4178 (1989 Jun)
Abstract
In human linkage analysis, many statistical problems without
analytical solution could be solved by ad hoc Monte Carlo procedures
were efficient computer-simulation methods available for members of
family pedigrees. In this paper, a general method is described for
randomly generating genotypes at one or more marker loci, given
observed phenotypes at loci linked among themselves and with the
markers. The method is based on a well-known expansion of the
multivariate probability of genotypes, given phenotypes, into a
product of conditional univariate probabilities that may be viewed
as corresponding to conditionally independent univariate random
variables. This representation allows a recursive evaluation of the
univariate probabilities that can be implemented in a surprisingly
simple manner by carrying out successive "risk calculations" with
respect to marker genotypes, given observed phenotypes and marker
genotypes already generated. Potential applications to various
unresolved problems are discussed. The method is applied to 28
published families analyzed for genetic linkage between hereditary
motor and sensory neuropathy I and the Duffy (FY) blood group locus
and confirms heterogeneity of hereditary motor and sensory
neuropathy I. An implementation of the simulation methods developed
in the LINKAGE program package will be available later in 1989.
