Information gain in joint linkage analysis.

The advent of DNA sequencing and the prospect of the availability of very large numbers of marker loci have reawakened interest in linkage studies. Statistical analyses of linked genetic loci have previously been based on pairwise analysis of the loci, and/or have assumed a known ordering of them along the chromosome. For the sizes of sample available in human medical genetics, pairwise analysis may be a statistically inefficient and inconclusive procedure. It is shown that three-way data provide very much more information, particularly with regard to the problem of ordering the loci. Correct answers to questions of locus order are of particular importance in current approaches to fetal diagnosis, where counselling is based on genotypes at markers closely linked to a disease locus. Although the three-locus analysis of this paper is only a first step towards joint analysis of multilocus data, the qualitative form of the data is of the joint pattern of cosegregation of loci. The pairwise counts of recombinants and nonrecombinants used for pairwise analysis are both qualitatively and quantitatively less informative. Many practical and theoretical problems of multilocus analysis of data remain to be solved, but the practical importance of doing so is demonstrated.