Radiation hybrid mapping has become an established tool for building p
hysical maps. It represents a powerful way of constructing YAC contigs
and high-resolution maps for positional cloning experiments. Ideally,
radiation hybrids should not only provide support for the true order
of the markers, but also accurate estimates of the physical distances
between them, Statistical analysis of radiation hybrids has proved dif
ficult because of the number of parameters (representing the fragment
retention probabilities) that must be estimated, and simplifying assum
ptions are needed to analyze large numbers of markers simultaneously.
The ramifications of these assumptions for the calculation of physical
distances are investigated. A simple two-locus model is presented to
demonstrate that variation in marker retention can lead to distortions
in the estimates of distance. Multilocus simulations show that, when
marker retention is constant. across the chromosome, good estimates of
physical distance can be derived using simple models of retention. Ho
wever, further simulations exploring variable retention schemes demons
trate that significant errors in the estimates of map distances can oc
cur. Ways of minimizing these distortions are discussed. (C) 1997 Acad
emic Press.