Predicting faces from genes

An article on GenomeWeb Daily News discusses some tantalising but still unpublished data from a team at Penn State University led by Mark Shriver:

The team has already found a handful of genes that seem to influence
different facial features. "I think we've got compelling evidence for
six genes that we tested," Shriver said. The team's data also suggests
that different parts of the face are influenced by different genes.

Variation in face shape is certainly strikingly heritable (all you need to do is look at any family photo to see that), but there's currently little known about the genetic architecture of that variation - in other words, whether it is determined primarily by just a few genetic variants of large effects (like skin colour), or by many genetic variants of small effect (like height).

Shriver's team seems to have focused primarily on the effects of genetic ancestry on face shape, using admixed populations such as African-Americans and Hispanics. In these samples there seems to be evidence for at least a few larger-effect variants (Shriver says, "Our initial results seem to show that not all of our genes have small effects.") However, obviously we'll have to wait until the paper comes out to see how compelling the associations are; and beyond that, I think a lot of people will be more interested in the genetics of within-population variation in face shape, i.e. why you look more similar to your immediate family than to your neighbours.

(As an aside, blogger Yann Klimentidis is a member of Shriver's team, and is also first author on a pertinent paper in PLoS ONE published this week.)

The GenomeWeb article notes the obvious potential application of these results in the forensic domain:

If things pan out the way [Shriver] thinks they will, genetic information from
a crime scene may provide information about the perpetrator's face that
could be used to prop up other evidence in court or provide useful
hints during investigations.

Exactly how accurate a DNA-based mugshot can be remains to be seen, but it seems likely that a combination of DNA-based ancestry estimates and physical trait prediction could provide at least vague but useful information for tracking down perpetrators.

Then of course there's the elephant in the room: pre-natal face prediction. I've previously described a likely future where couples undergoing IVF select which embryos to implant based on a table of risk factors and probabilistic trait predictions for each embryo, derived from large-scale genetic testing. Imagine if those parents could also see a visualisation of the probable facial structure of each of their potential children at the same time (or, more likely a "cloud" of potential faces indicating the uncertainty of the prediction).

Would this be a popular option, or would it simply be too uncomfortable for parents-to-be to stare into the potential eyes of each ultimately discarded embryo?

A somewhat less ambiguously happy scenario is the use of fetal DNA extracted from a pregnant woman's blood to create a fuzzy prediction of the adult face of the child-to-be. Given the extraordinary appeal of things like four-dimensional ultrasound scans of developing fetuses, it's easy to see that such a technique would have impressive commercial appeal for expecting couples.

Subscribe to Genetic Future.