The maximum LOD score of the NPL linkage analysis was 3.75 (P = 1.6 × 10-5
) at 110.5 cM (corresponding to 63.4 Mb in physical distance), when the marker-marker LD was modeled at either r2
> 0.5 or r2
> 0.2. This result supported significant evidence for linkage between the 19q region and COME/ROM (Figure ). The LOD-1 support interval (the interval in which the LOD score is within 1 unit of its maximum, which is usually treated as a confidence region [30
]) for the identified linkage was between 107.3 cM and 111.1 cM (61.6 Mb and 63.8 Mb in physical distance). When the marker-marker LD was not modeled, the maximum LOD score was 4.41 at 109.4 cM. The large difference of the maximum LOD scores between the two models (with LD modeled vs. not modeled) reflects the inflation of linkage due to the unaccounted marker-marker LD among the dense SNPs [22
Figure 1 LOD scores at 19q. The dotted line indicates a non-parametric linkage scan without modeling marker-marker LD. The dashed and solid lines indicate non-parametric LOD scores with LD modeled at r2 > 0.5 and r2 > 0.2 respectively. The dot-dash (more ...)
When 90 families (65% of total number of families) with complete parental data were included in the linkage analysis, the maximum LOD score was 2.85. Note the other 35% of data contains additional linkage evidence that can be estimated as shown above [22
]. Since the marker-marker LD does not influence the linkage analysis in 65% of the data even without modeling LD in the analysis [22
], and our result provides a projected LOD score of 4.38 for the entire set of families for the hypothetical case when all 138 families have complete parental information, the identified linkage in the 19q region is further supported.
The maximum LOD score based upon parametric linkage analysis was 5.42 at 107.5 cM, corresponding to P = 1.6 × 10-5. For a sibling of an affected individual, the recurrence risk of COME/ROM attributable to the 19q locus is twice as high as the recurrence risk in the population.
No statistically significant (P < 0.0003) association between any SNP and COME/ROM was identified (Figure ). Using a Bonferroni correction, 1,091 SNPs requires a nominal significance level of P < 4.6 × 10-5. Since the number of independent SNPs should be less than 1,091 because of the LD between SNPs, we estimated the number of independent SNPs by pruning pairwise LD between SNPs. There are 216 SNPs that are in approximate linkage disequilibrium with each other (r2 < 0.1), and a Bonferroni correction based on 216 SNPs gives a nominal significance level of P < 0.00023. None of the observed associations reach either level of significance.
P values of association from two association tests: A. TDT, B. GDT, C. MQLS. P values in -log10 scale are represented by circles. The solid line indicates p values of non-parametric LOD scores with LD modeled at r2 > 0.2.
The five strongest associations identified through the TDT family-based association analyses (P value < 0.001) all fall in the LOD-1 support interval. In contrast, none of the family-based GDT or MQLS
analyses yielded significant evidence of association (P < 0.001). The strongest association detected by TDT reflects the inflation of the statistics due to linkage in the region, rather than by association at a single variant [25
]. Using GDT analysis, the strongest association (P = 0.0024) falls outside of the LOD-1 support interval.
Rare variants in this region of 19q are much less represented than common variants in our SNP panel. Among all 1,091 SNPs, there are only 81 SNPs with a minor allele frequency less than 5%. The lack of significant association is consistent with evidence that linkage in the 19q region is not due to common genetic variants.