Faces are everywhere in Mark Shriver's office. African face masks are tacked to the wall, 3-D printed faces of himself and colleagues hang nearby, and family photos are taped near his desk.
Humans are naturally attracted to faces. Mere hours after birth, newborn babies show preference for faces over any other pattern. And this predisposition makes sense. After all, your face is one of the first things other people notice about you. It's the most recognizable part of your identity.
Shriver, Penn State professor of anthropology, has turned his own fascination with faces into a major part of his career. In conjunction with Peter Claes, anthropology professor at Katholieke University of Leuven in Belgium, Shriver has been using 3-D modeling, a computer program and a person's DNA to predict what his or her face looks like.
As the head of Penn State's Anthropological Genomics Lab, Shriver studies how and why humans have evolved to look so different from one another. And although he's also researched several traits that vary across populations -- like skin color and voice pitch -- Shriver has recently been focusing on the evolution of the face.
"I really want to understand the evolutionary process and how faces have been affected by evolution," said Shriver. "The variation in people's faces -- how different we all look -- it almost seems excessive. So there's probably a reason for that, and I'd like to better understand it."
Predicting someone's face from their DNA is a two-step process that uses a computer program developed by Shriver and Claes called FacePred. First, Shriver begins with a "base face," which is an average face based on sex and genomic ancestry -- for example, a South Korean woman. Then, he uses the person's DNA to overlay the changes in the person's facial features based on their own genetic marker profiles, known as genotypes. The final predicted face can then be compared to the person's actual face to see how accurate the prediction was.
The results vary. Some of the predicted faces look very close to the person's real face and others miss the mark in places. Shriver says he and his colleagues are currently working to improve the system, creating more accurate base faces and learning more about genetic markers.
"We're working to screen more markers and advance the algorithm that predicts the faces," said Shriver. "So things are progressing nicely, although we do need to collect many more sample faces."
But DNA is just one part of the puzzle. Although a person's genetic makeup largely determines what his or her face looks like, environmental factors -- like sun exposure or whether a person smokes or not -- also have a role in facial appearance. A lot of time basking in the sun might change the skin's texture, for example, and habitual smoking can add wrinkles.
Julie White, a Penn State student working toward her Ph.D. in anthropology and a member of Shriver's lab, wants to discover just how much of our facial appearance is "heritable" -- or how much comes from our DNA and how much comes from other factors.
"We all know that facial features run in families. I have my father's nose, for example, and my sister and I have near identical face shapes," said White. "But we don't know exactly how it works. We can't yet say, for example, 'Oh, 65 percent of a face's variability is genetic and 35 percent is environmental.' But I'm hoping to figure that out."
To help her better understand the face's heritability, White traveled with Shriver and the other members of his lab to the Twins Day festival in Twinsburg, Ohio, during the second weekend in August. There, they set up a tent and took 210 3-D facial images, including those of 77 sets of identical and fraternal twins.
The images, taken by a special 3-D camera, will help White learn more about how genetics affect a person's face. Since identical twins have identical DNA, White says it's generally safe to assume that any small differences in the face -- a crooked nose, perhaps -- come down to environmental factors. She'll also look at images of fraternal -- or nonidentical -- twins, who on average share about 50 percent of the same DNA.
"As a hypothetical example, if facial features are 100 percent heritable, meaning that 100 percent of the variation in faces is due to genetic factors, then we would expect the faces of the identical twins to be exactly the same, and the faces of the nonidentical twins to be 50 percent alike," said White. "But we already know that facial variation and facial features aren't completely due to genetic factors because even identical twins have slight differences. So, the name of the game is to figure out how much facial variation is influenced by genetics and how much is influenced by the environment."
By comparing many sets of twins, White says she hopes to draw conclusions about how much of your face's appearance comes down to genetics.
Being able to better understand heritability and predict faces from DNA could play an important role in solving crimes. If a criminal leaves DNA at the scene of a crime, predicting what his or her face looks like could help whittle down a list of suspects. Or, scientists could predict what the faces of Egyptian mummies or frozen Neanderthals looked like, giving us key insights into our past.
The technique could also help predict the future.
One of the photos taped up in Shriver's office is a snapshot of his son, Hector. Giving the camera a mischievous grin, the nine-month-old flashes two tiny teeth. Although it's clear he's adorable now, Shriver was curious about what he would look like in the future. So Shriver had him genotyped and was able to predict what his face will look when he's 25 years old.
Shriver has that picture in his office, too. But seeing as Hector is now barely two years old, it will take much longer to determine how accurate that prediction is.
For more IT stories at Penn State, visit http://news.it.psu.edu.