Some developments on the affected-pedigree-member method of linkage analysis

Some improvements are presented for the affected-pedigree-member method of linkage analysis, which is a generalization of the sib-pair method. The test statistic is extended to include contrasts between affected and unaffected pedigree members, so that it now utilizes marker information from all typed pedigree members rather than just the typed affected members. Computer simulation using a sample pedigree of 14 individuals shows that this modification can substantially increase statistical power where there is a direct association between marker variation and disease and where disease risk is elevated in carriers of the disease allele. Data on Huntington disease in 16 British families, which were analyzed previously using only the affected individuals, are reanalyzed with the unaffected individuals included. Strong rejection of the null hypothesis of no association between Huntington disease and the HindIII polymorphism is confirmed, but the particular families in which the association is significant differs from that obtained through an analysis based only on affected individuals and reflects more closely the results obtained from a lod-score analysis. The test statistic is also modified here to incorporate contrasts between individuals of zero kinship, if needed. This enables contrasts between individuals from different pedigrees, as well as contrasts involving individuals sampled from the general population, to be incorporated into the test of association. For population data, the methodology reduces to a type of contingency-table analysis, in which the rows of the table correspond to different marker-locus genotypes and in which the two columns categorize subjects into an "affected" group versus an "unaffected," or control, group. This aspect of the methodology is illustrated using two population data sets, the first relating APO-E genotype to the frequency of individuals undergoing maintenance hemodialysis and the second relating APO-B genotype to the frequency of coronary artery disease. The present methodology confirms the lack of association between marker and disease in the former data set and confirms the presence of association in the latter. Finally, the methodology is formulated here in terms of ordinary, multiperson kinship coefficients rather than in terms of the generalized kinship coefficients originally proposed. This greatly reduces the number of coefficients to be calculated, thereby enhancing the computational efficiency of the computer program.