A penalised logistic regression model was created for blonde colour, including all lead variants for blonde vs. black and brown hair colour.
But as the couple grow up and grow apart, Emma lets the blue-dye job grow out and she reverts to her natural blonde colour.
Dye your hair a light blonde colour and put it up in the wavy, high-fun ponytail style.
She now has a range of blonde colours for hair as well as brunette and auburn.
The regularized effect estimates were used to create genetic scores for blonde hair colour, assuming an additive genetic model.
Whilst red hair is essentially a Mendelian trait modified by additional loci, the genetic architecture of blonde hair colour is concordant with a polygenic trait.
There were just three SNPs that were ~10−8 in blonde hair colour, and only one extra round of conditional analysis was needed.
Following conditional association testing to uncover additional signals of association, we discover 213 lead variants associated with blonde hair colour (Fig. 1b, Supplementary Table 4, Supplementary Figure 1b).
Specifically we constructed a blonde hair colour polygenic phenotype score by taking the variants that reached genome-wide significance in the blonde vs. brown and black hair colour conditional analysis (5×10−8), as a linear combination of the allele-weighted regularised logistic regression coefficients.
Remarkably, we have identified seven of these, ERRFI1, FRAS1, HOXC13, PADI3, KRTAP, PEX14 and LGR4 as affecting blonde/non-blonde hair colour (P = 1×10−11, Fisher's exact test).
3 Light skin colour, blonde or red hair, number of moles, 4 family history 5 and excess exposure to ultraviolet (UV) radiation, particularly in childhood, are the main risk factors for malignant melanoma.
