In linkage analysis with affected related pairs (ARP), stated familial relationships are usually assumed to be correct, thus misspecified relationships can lead to either reduced power or false-positive evidence for linkage. In practice, studies either discard individuals with erroneous relationships or use the best possible alternative pedigree structure. We have developed several linkage statistics that model the relationship uncertainty by properly weighting over possible true relationships. We consider ARP data for a genome-wide linkage scan. A simulation study is performed to assess the proposed statistics, and to compare them to the maximum likelihood statistic (MLS) and S_all LOD score using true and discarded structures. We have simulated small and large pedigree datasets with different underlying true and apparent relationships, and typed for 367 microsatellite markers. The results show that two of our relationship uncertainty linkage statistics (RULS) have power almost as high as MLS and S_all using the true structure. Also, these two RULS have greater power to detect linkage than MLS and S_all using the discarded structure. Thus, our RULS provide a statistically sound and powerful approach for dealing with the commonly encountered problem of relationship errors. The RULS are relevant to public health because application of these RULS to complex human disease will facilitate the mapping and discovery of genes involved in the etiology of such diseases.

We attempted to apply RULS to Otitis Media with effusion (OME) data from Caucasian families. OME is an infection causing fluid in the middle ear, and is the most common cause of hearing loss among young children. We have recruited subjects (with history of tympanostomy tube insertion) and their families (parents and affected/unaffected siblings). Genotyping was done using Affymetrix 10K SNP chips, and out of 1,584 enrolled individuals (322 families), 1,191 (305 families) are genotyped at this date. We performed nonparametric multipoint linkage analysis using discarded structures. The preliminary results show suggestive linkage peaks on six chromosomes, the highest being at rs1345938 on chromosome 7 with S_all LOD score of 2.36 (p-value 0.0005).