We are requesting the P values for each marker from genomewide analysis of available samples in a number of ALSPAC phenotypes. We need the genome-wide set of P-values to compare with a similar set of measures obtained in a study of mouse complex phenotypes. Our justification for wanting to carry out this comparison is as follows:

Genetic mapping of disease models in rodents has long been expected to help with the identification of genes involved in human illness. Yet, while this is true for models that arise from abnormalities of a single gene, leading to new insights into diverse conditions such as obesity (leptin), susceptibility to infectious disease (toll receptors) and mental retardation (lissencephaly), the approach has been far less successful for phenotypes with a complex genetic basis. In a few instances common genes have been found: variants in CTLA4 increases risk of autoimmune disease in humans and a mouse model of type 1 diabetes (Ueda et al. 2003); a copy number variant in the same susceptibility gene (Fcgr3) contributes to immunologically mediated glomerulonephritis in humans and rats (Aitman et al. 2006).

The conservation of many physiological processes between the two species suggests that variation at genetic loci might be shared, and there is some evidence that this is so. Paigen and colleagues used linkage results to argue that more than half of human atherosclerosis QTLs are located in regions homologous to mouse QTLs (Wang et al. 2005a). However human linkage results are notoriously unreliable and, as Risch and colleagues pointed out, concordance is unlikely given that the genetic effect in humans depends on disease allele frequencies and such allele frequencies are unpredictable (Risch et al. 1993). The lack of power in human linkage studies compromises their use in comparison with the mouse findings. Furthermore, while the mapping of complex traits has proved to be an effective technique, delivering many thousands of quantitative trait loci (QTLs) its power is not matched by high resolution. The large confidence intervals into which both mouse and human QTLs were mapped made it difficult to exclude the possibility that overlap was coincidental.

Two developments allow us to revisit the question of the overlap between mouse and human QTLs, First, the advent of adequately powered genome-wide association studies in humans, which map genetic effects at high resolution, has proved to be a robust method for QTL identification (McCarthy et al 2007, Nature Reviews Genetics) Second, progress in mapping QTLs at high resolution in mice has yielded a map of over 900 QTLs for 100 phenotypes, each mapped to approximately 3 Mb (Valdar et al. 2006a). This was achieved using a quasi outbred population of mice, the heterogeneous stock (HS). While HS mouse QTLs are not mapped to the same resolution as is obtainable in human association studies, we hypothesize that it should be possible to perceive overlap between data sets.

In order to test this idea, we need a large set of data comparable to that mapped in the HS. All of our data, genotypes and phenotypes are freely available online (http://gscan.well.ox.ac.uk). We wish to compare the distribution of the mouse QTLs to the distribution of P-values found in human GWAS studies. Our mouse data set contains the following phenotypes that we wish to compare with the GWAS data from ALSPAC, Table 1 (appendix).

To carry out our analyses we need the P values for each marker. We do not need the individual genotypes, nor the individual phenotypes. We need the result of the analysis of association for every marker in the genome.