The reference family panel is the foundation of a gene mapping program beca
use it affects the cost and quality of the genetic linkage maps, and should
be designed to yield reliable linkage detection and locus ordering at mini
mal gene mapping cost. A map cost function was defined as the number of gen
otypes required per marker per unit of genome coverage and was used to obta
in optimal designs with respect to linkage detection. An ordering reliabili
ty function was defined as the likelihood ratio of the most likely order to
the second most likely order of genetic markers and was used to find optim
al designs with respect to locus ordering. Optimum levels of recombination
frequency were found to be in the neighborhood of 0.11-0.15 for linkage det
ection and were in the region of 0.05-0.20 for locus ordering. Therefore, r
ecombination frequencies optimal for linkage detection are also optimal for
locus ordering. Based on the optimal detection levels, sample size (number
of offspring) and map cost requirements were derived for six representativ
e designs, assuming gender-specific linkage maps and two alleles with equal
frequency for each marker. The sample size required for linkage detection
ranged from 168 to 432 offspring for full-sib designs and ranged from 350 t
o 600 offspring for half-sib designs depending on the family size and the t
arget LOD score, with corresponding minimal map costs of 10-20 genotypes pe
r marker per centiMorgan map coverage. Locus ordering generally requires mo
re genotypes than linkage detection. For full-sib designs, meioses from bot
h genders should be used for locus ordering even when the maps are gender-s
pecific. For half-sib designs, additional families may be needed for locus
ordering. Sample size for ordering closely linked loci as required by posit
ional cloning were provided. Effects of family size, grandparents, and mark
er polymorphism on design efficiency were analyzed.