A test of nonrandom segregation

Within a family, associations between a disease and a marker locus are often inferred when affected offspring share marker alleles more often than is expected by chance. Generally, this is due to nonrandom parental transmission of marker alleles and specifically could be due to linkage, epistatic gene action, or segregation distortion at the marker locus. In this paper, we discuss the statistical properties of a general test of nonrandom segregation of a marker gene. The exact probability distribution of the test under the null hypothesis of random segregation is derived, as is the distribution under the alternative hypothesis of genetic linkage. We compute the mean and variance of these distributions as a means of judging the adequacy of random segregation to explain disease-marker data but also provide a method for computing the exact significance value under the null hypothesis. These methods have been utilized for studying HLA segregation in families with tuberculoid leprosy. On the assumption that this type of leprosy is autosomal recessive, we find evidence that a gene controlling susceptibility to infection by Mycobacterium leprae resides on human chromosome 6, approximately 13 map units away from the HLA locus in males.