It is sometimes forgotten that linkage studies provide information about rare, relatively penetrant susceptibility loci. Family-based designs are typically not well powered to detect FRAX597 the small effects found in GWASs. For example, on average, siblings share 50% of their genome. Where two siblings have the same disease, departure from this 50% sharing indicates regions

that harbor risk variants; but since the SD for sharing is large (approximately 3.7%), large sample sizes are required to detect a significant departure. Family designs can however detect one form of genetic variation that is hidden from GWASs: the joint effect of independent, rare, mutations in the same gene (recall that GWASs are effective for common variants). In a linkage study, the effects of Roxadustat mw independent mutations will combine together, since the unit of analysis in linkage (the average distance between recombinations in the human genome in a single meiosis) is a much larger genomic region than is the case for association analyses. In cases in which linkage asserts that there is an effect but association fails to detect one, then one explanation is allelic heterogeneity: multiple effects exist in the gene but on different haplotypes. Linkage studies are summarized in Table 3. Results are reported as a logarithm of the odds (LOD) score, rather than

a p value. The majority of the studies reported in Table 3 used an affected sibling design tuclazepam (in which two siblings have MD). In this design, an LOD score of 2.2 is suggestive evidence for linkage (expected to occur once by chance in a genome scan), an LOD score greater than 3.6 represents significant linkage (expected to occur by chance with a probability of 5%), and an LOD score of 5.4 is highly significant (probability of chance occurrence is less than

0.1%) (Lander and Kruglyak, 1995). Table 3 makes four points. First, there is clear heterogeneity between studies. The outlier here is the Zubenko study (Zubenko et al., 2003), which reports more loci at higher levels of significance than all the others. Second, there is evidence for poor internal consistency. Three groups report data in multiple publications, usually because they acquired additional data (Utah families [Abkevich et al., 2003 and Camp et al., 2005], DeNt [Breen et al., 2011 and McGuffin et al., 2005], and GenRED [Holmans et al., 2004, Holmans et al., 2007 and Levinson et al., 2007]). The additional samples collected by the GenRED consortium failed to confirm the 15q linkage reported in their initial paper (Holmans et al., 2004). The authors considered that the first finding might be a false positive, that the second finding might be a false negative, or that both findings were true, the difference being attributable to variation in the clinical features of the families (Holmans et al., 2007). Third, there are overlaps in the locations identified by linkage results (Table 3).