The nonsynonymous nucleotide position rs1800407 (exon 13) was found to be associated with both eye colour (p = 0.001) and skin type (p = 0.015). In our data, this SNP was associated with eye colour in the blue versus nonblue test, but the most significant association was found with green/hazel iris colour (p < 0.001). These results were also statistically significant after adjustment for multiple tests. The second studied nonsynonymous polymorphism (rs1800401) was found not to be associated with eye colour in the blue versus nonblue test. An additional test revealed significant association for this SNP in the blue/green versus hazel/brown test (p = 0.035). This result is, however, not significant after Bonferroni's multiple-comparison correction.
In order to eliminate the problems of statistical dependence among analysed SNPs, which, as we found, revealed extensive linkage disequilibrium, the data were subjected to haplotype reconstruction. Using a Bayesian procedure implemented in the PHASE program, twenty nine different haplotypes were inferred. Table 3 presents relevant haplotypes associated with phenotypic effect according to stratified analysis or tree scanning. The stratified analysis revealed three haplotypes (4, 6, 27) to be significantly associated with nonblue eye colour and another one (21) with blue irises. The most significant association was noted for haplotype 6 (p = 0.0015), which was observed 46 times. As haplotypes are themselves correlated due to a common evolutionary history, it is assumed that consideration of this evolutionary data may improve the effectiveness of association studies (Posada et al. 2005; Templeton, 1995). Thus, the inferred haplotypic data were subjected to the method developed by Templeton et al. which assumes genealogical relationships of haplotypes reflecting their evolutionary history (Templeton et al. 2005). Since the procedure implies analysis of a phylogenetic tree of haplotypes, the reconstructed haplotypes were first used for inference of evolutionary trees using the parsimony approach implemented in the DNAPARS computer program (Felsenstein, 2005). As shown in Table 2, two very close positions located in intron 16 were analysed in this study. However, application of both was advantageous for haplotype inference and further phylogenetic analyses, as the omitted AY392134 resulted in 28 instead of 29 reconstructed haplotypes, which when subjected to phylogenetic tree reconstruction resulted in a very high level of ambiguity as reflected by 1206 most parsimonious phylogenetic trees. An inference performed on the basis of 10 polymorphisms revealed phylogenetic ambiguity which could have been resolved into 32 most parsimonious evolutionary trees of various topologies. All phylogenetic trees were then analysed with the Treescan program (Posada et al. 2005). The scanning was performed for three different pigment characteristics – hair colour, skin phenotype and eye colour. Table 4 shows the results of the first round of the tree scanning test obtained for eye colour, the only trait associated with the studied polymorphisms according to this method. The different values presented in this table for each of the analysed factors reflect the various results noted for different trees. The initial scan based upon 1000 permutations revealed associations in the case of eye colour (p≤ 0.05) sixty four times linked with four different branches, sixteen times respectively for branches 7–8, 8–6, 7–5, 5–6 (see Fig. 1). To obtain the second round of scanning all detections from the first round of analysis are taken into account, thus the second round is simply conditioned on the branches detected during the first round. The second round revealed associations 28 times always with branch 5–6 (conditioned on 7–5 – 15 times, on 8–6 – 8 times and 7–8 – 5 times, see Table 5 and Fig. 2). The results obtained clearly indicate that the strongest phenotypic effect is associated with branch 5–6. Careful inspection of haplotypes 5 and 6 reveals that branch 5–6 is defined by rs1800407, i.e. an amino acid causing change Arg419Gln located in exon 13 (Table 3). Interestingly, branch 7–8 associated with eye colour in the initial round of tree scanning is also defined by this same polymorphism. It is worth noting that either branch 5–6 or 7–8 or both of them were detected for all 32 scanned most probable evolutionary trees. It also seems possible that in the case of detected branches 7–5 and 8–6, both defined by rs1800404 (synonymous change in exon 10) the true phenotypic association may be actually due to branch 5–6, but tree scanning assigned the association to an adjacent branch due to some stochastic fluctuations. The consensus network constructed using the SplitsTree4 computer program (Huson & Bryant, 2006) on the basis of all 32 most parsimonious trees shows ambiguity associated with detected branches 5–6, 7–8, 8–6 and 7–5 (Figs 1 and 2). The second step analysis conditioned on branch 7–5 indicated branch 5–6 as associated with eye colour in all cases but one. Similarly, in the case of initial branch detection 8–6, the second round indicated branch 5–6 in 50% of cases. Figure 3 presents the haplotype tree for which most significant associations were found. Neither the stratified analysis nor the tree scanning revealed any trace of association with hair colour or skin type for the analysed population sample.