Overview DHSMAP is a program, written in C++, for fine-mapping of qualitative traits by linkage disequilibrium. The software is based on the Decay of Haplotype Sharing (DHS) method for fine-scale linkage disequilibrium mapping by McPeek and Strahs (1999). Given a set of marker haplotypes or genotypes from affected individuals, haplotypes or genotypes from appropriately selected controls, and a genetic map of the markers at which both sets of individuals are typed, DHSMAP estimates the location of the trait-associated variant by maximum likelihood or maximum quasi-likelihood. (Note that, for the controls, it is possible to use marker allele frequencies as input to the program instead of haplotypes or genotypes, but this is not recommended except in situations in which there is known to be no background linkage disequilibrium; see "Tips" for more information.) Consider an ancestral haplotype on which a variant is introduced. DHS models the decay of sharing of the ancestral haplotype by descendents, where the number of haplotypes sharing ancestral DNA decreases with increasing genetic distance from the variant. Heterogeneity in ancestral haplotype is permitted. The DHS model includes parameters 1/tau (expected genetic distance to which the ancestral haplotype is preserved), a heterogeneity parameter p (proportion of variant haplotypes NOT descended from the ancestral haplotype), the ancestral haplotype of marker alleles, and the location of the variant. The DHS model also allows for mutation and chance sharing of sections of marker haplotypes. In the current implementation, the mutation rate may vary across loci, but when a mutation occurs, it is equally likely to change to any of the other alleles at the locus. In allowing for chance sharing, DHSMAP models the probability of a haplotype drawn from the control population using a 1st or 2nd order Markov model to take into account background linkage disequilibrium; the probability of observing a particular marker allele at a locus depends on the adjacent allele. DHSMAP estimates the required conditional allele frequencies, i.e., the frequency of an allele at a locus given a particular allele appears at the adjacent locus, from the control haplotypes or genotypes by maximum likelihood using a hidden Markov model (HMM). Conditional on the location of the trait-associated variant and on the ancestral haplotype, the likelihood and quasi-likelihood of the affecteds' haplotype or genotype data is calculated and maximized over the parameters 1/tau and p using a HMM in which the embedded Markov chain models the sharing by descent from the ancestral haplotype, while the observations are the marker alleles. DHSMAP computes the likelihood across individuals assuming independence of the recombinational histories of the descendent haplotypes since the ancestral haplotype and also calculates the more conservative quasi-likelihood assuming a conditional coalescent genealogy. We use a 3-step procedure to maximize the likelihood and quasi-likelihood over the ancestral haplotype and the location of the variant (See "Search Procedures" for description and "Tips" for advice). The location with the highest maximized likelihood or quasi-likelihood is DHSMAP's estimate of the variant's position. DHSMAP generates approximate 95% confidence intervals, for the likelihood and quasi-likelihood estimates, by inverting the likelihood ratio test as described in McPeek and Strahs (1999).