Two-locus likelihood

Jurg Ott OTTSR%CUVMB at CUVMB.CC.COLUMBIA.EDU
Mon Aug 16 07:59:05 EST 1993


Martin Farrall asks:

     Is it fair to say that HOMOG3R which to quote from the
     documentation that accompanies it:
          "This is a specialized version of the HOMOG3 program.
          It calculates log likelihoods under the assumption that
          a trait is linked in a proportion alpha1 of families to
          marker 1 in some area of the genome, and in other
          families it is linked to a marker 2 in another area of
          the genome."
     computes a joint lodscore that is comparable with that
     computed under the "additive" model of Risch (see Am J Hum
     Genet 46:222-253)?

Here is how I see things: For rare traits the two models (hetero-
geneity as implemented in HOMOG3R, and Risch's additive two-locus
model) are equivalent and, thus, will produce comparable likeli-
hoods.  The major difference between the two models is that in
the heterogeneity approach one allows for heterogeneity only
between families but assumes homogeneity within each family.  For
rare traits this assumption is reasonable, but for the more
common traits it may well occur that disease genes enter a family
more than once ("bilineal" families).  It was for the purpose of
handling this situation in common traits that we started imple-
menting two-locus disease models in the LINKAGE programs (Lathrop
and Ott, Am J Hum Genet 47:A188, 1990).




More information about the Gen-link mailing list