Page 192 - Applied Probability
P. 192

177
                                                                  9. Descent Graph Methods
                              of a prior probability and a transmission probability.
                                Under the usual assumptions of genetic equilibrium, Prior(G) is the prod-
                              uct of the population frequencies of the founder alleles involved in G. Since
                              a descent state entails no ambiguities about recombination, Trans(G) re-
                              duces under Haldane’s model of recombination to a product of a power of
                                                                                              2 1
                              and relevant powers of the recombination fractions and their complements
                              for the adjacent intervals separating the markers. Finally, owing to the fact
                              that all compatible descent states G  ø  G exhibit the same transmission
                                                                   0
                              pattern, we can reexpress the likelihood (9.8) as

                                           Pr(G ∩ M)   =  Trans(G)        Prior(G).        (9.9)
                                              0
                                                                0
                                                                   G  à  0 G∩M
                              In the next section we tackle the subtle problem of quick computation of
                              the sum of priors         Prior(G) [19, 34].
                                                G  à  0 G∩M
                              9.5 Descent Trees and the Founder Tree Graph
                              Given l loci in a pedigree with p people and f founders, there are 2lp nodes
                              in a descent graph. These nodes are grouped in 2lf descent trees. The
                              descent tree rooted at a particular founder node contains that founder node
                              and those non-founder nodes inheriting the corresponding founder gene. All
                              nodes of a descent tree involve the same locus. When a founder gene is not
                              passed to any descendant of the founder, then the descent tree exists but
                              is degenerate.
                                It is convenient to proceed to a higher level of abstraction and make
                              the founder trees into an undirected graph. This abstraction serves to keep
                              track of how founder alleles are constrained in a coupled manner by the ob-
                              served marker phenotypes in the pedigree. The nodes of the founder tree
                              graph are the descent trees of the descent graph. Two nodes of the founder
                              tree graph are connected by an edge if and only if the two corresponding
                              descent trees pass through the same typed locus of some person in the
                              pedigree. This definition precludes connecting two descent trees associated
                              with different loci. Part (a) of Figure 9.2 shows a descent graph for a single
                              marker locus in which each descent tree is labeled above its rooting founder
                              gene. Do not confuse these labels with the allele symbols used in descent
                              states. Part (b) of Figure 9.2 shows the founder tree graph corresponding
                              to this descent graph, assuming all nonfounders and no founders are typed
                              at the locus.
                                It is possible for two descent trees at the same locus to mutually impinge
                              on more than one person typed at the locus. Although this information
                              is relevant to discerning whether the two trees are genetically compatible
                              with the observed phenotypes in the pedigree, for the sake of simplicity, we
                              will still view the descent trees as connected by just a single edge. When
   187   188   189   190   191   192   193   194   195   196   197