Fan Liu, Kate van Duijn, Johannes R. Vingerling, Albert Hofman, AndrÃ© G. Uitterlinden, A. Cecile J.W. Janssens, Manfred Kayser (2009). Eye color and the prediction of complex phenotypes from genotypes Current Biology, 19 (5) DOI: 10.1016/j.cub.2009.01.027
In a recent post I noted that genetic tests to predict adult height are still a long way off being accurate; currently, known genetic variants can predict just over 5% of the variance in height, as opposed to 40% predicted using a simple algorithm based on the heights of both parents. The genetic complexity of height means that trying to screen embryos for this trait using pre-implantation genetic diagnosis is likely to be little more than an exercise in frustration.
However, that's not true for all traits. In several recent posts I've mentioned eye colour as one relatively genetically simple trait that would prove amenable to embryo screening, and indeed there has already been at least one US fertility clinic offering such screening to couples undergoing IVF (although that offer has now been withdrawn).
A paper in the most recent issue of Current Biology puts some numbers on the predictive value of genetic testing for the prediction of eye colour. The authors examined all of the 37 variants (in 8 separate genes) previously reported to have an association with eye colour in a sample of 6,168 Dutch of European ancestry, and assessed the predictive value of these markers alone and in combination.
The results were clear: with just six genetic markers, individuals with blue or brown eye colour could be predicted with over 90% accuracy*; the accuracy for predicting intermediate colours (e.g. green) was somewhat lower at around 72%.
Adding further markers had rapidly diminishing returns, with the last 15 markers adding essentially nothing further in terms of predictive value (these markers were typically captured by other variants within the same gene).
The authors note that "[t]he genetic prediction values obtained here for blue and brown eyes in
Europeans represent the highest accuracies revealed so far in genetic
prediction of human complex phenotypes." This makes for a pleasant change from the results for most disease-related traits, where common genetic variants have generally provided frustratingly little predictive power for risk prediction.
In addition to being a tempting target for "cosmetic" genetic screening by parents undergoing IVF, eye colour genes will no doubt prove useful in forensic applications - being able to predict any physical traits from trace DNA left at the scene of a crime will at least occasionally be useful for investigators. However, the researchers note that their results should only be regarded as applying to individuals of European ancestry; different populations are known to have different genetic determinants of pigmentation.
* Strictly speaking, the researchers showed that the area under the ROC curve was greater than 0.9 for blue and brown eyes and 0.72 for green eyes; this can be interpreted as the probability that a randomly chosen positive instance would be ranked higher than a randomly chosen negative one.
I believe that the technology will never reach 100%. There will always exist a small chance that such predictors and markers just happen to fall into the statistically extreme area, so we can never guarantee a prediction of eye color. However, eye color doesn't matter much anyways, so let parents predict all they want. It has no tangency with intelligence or physical superiority.