Defining a Face: What Can DNA Phenotyping Really Tell Us About An Unknown Sample?

by Jim Dawson

September 2015

Within every person, somewhere among the approximately three billion DNA base pairs, hidden in the alleles and single nucleotide polymorphisms, is the information that defines much of an individual’s physical appearance. This DNA-determined appearance, or phenotype, is what creates family resemblance and, in the words of geneticist Richard Spritz, is “what your grandmother is responding to when she says you look like your father.” Efforts by geneticists to find the pieces of DNA that determine what a human face looks like—everything from the shape of the nose to the spacing between the eyes—have intensified in recent years, and progress has been made. Scientists can now, with some certainty, use a strand of DNA to identify an individual’s likely hair and eye color, as well as skin pigmentation and ancestry. Penn State University geneticist Mark Shriver has made what he describes as the “first effort at generating facial composites from DNA” with “preliminary but certainly promising results.”

Forensic R&D at NIJ

NIJ is the federal government’s lead agency for forensic science research and development. Our mission is to improve the quality and practice of forensic science through innovative solutions that support research, development, testing, and evaluation.

This article was authored under a contract from NIJ. It was originally published in Forensics Magazine and is reprinted with permission.

Creating a “photo” image of an individual’s face from strands of DNA is of enormous interest to forensic investigators. A physical portrait of a suspect might be developed from DNA left at the scene when there were no eyewitnesses to the crime. In the aftermath of fires or other catastrophic events, DNA from unidentifiable bodies could make them recognizable to family members. In addition, DNA from a bone fragment could help visualize and identify individuals in mass graves. Although most geneticists stress the limits of phenotype research, police and forensic investigators can already turn to a handful of private companies that claim they are able to use DNA to accurately predict an individual’s physical appearance. So, what is the state of the phenotype research? Can a forensic investigator realistically expect to get an accurate image of an individual from a piece of DNA?

“We are not even at the end of the beginning,” said Spritz, who has spent six years trying to identify and understand what determines the appearance of a human face. Spritz, the program director of the University of Colorado’s Human Medical and Genetics Program, is one of several DNA phenotype researchers receiving support from the National Institute of Justice. Although geneticists are cautious about overselling the progress toward creating an accurate physical image of an individual from DNA, there is a general agreement that the understanding of the underpinnings of phenotypes has moved forward significantly in the past decade. Scientists can now use DNA to determine, with more than about 75 percent probability, an individual’s ancestry, eye and hair color.

Much of the ancestry work is being done by Yale University geneticist Kenneth Kidd, who has developed a panel of 55 “ancestry informative single nucleotide polymorphisms” (AISNPs), which divide people into eight geographical regions, such as Europe, East Asia, and the Pacific. DNA from a bone fragment found in Vietnam, for example, can be screened against the AISNPs panel to determine if the person was from Southeast Asia or North America. If that person was an African-American, however, the results would come back as Ethiopian because that is a mix of European and African genes. Kidd is expanding the AISNPs panel to include more geographical regions. Identifying an individual’s ancestry is a piece in the genetic puzzle that determines what that person looks like, and it is an important part of the broader effort to accurately portray a specific face from DNA. Despite the progress he is making, Kidd said he still has a long way to go. “With the sort of research I’m doing,” he said, “I’ll never be finished.”

More than just ‘brown vs. blue’

At a research lab at Indiana University–Purdue University Indianapolis, geneticist Susan Walsh is working to refine DNA phenotyping to predict quantitative color—or the precise color of eyes, hair and skin. Earlier work by Walsh and others identified the single-nucleotide polymorphisms, or SNPs, that drive pigmentation. “That is the categorical identification,” she said, “brown versus blue eyes, blonde versus brown hair. Our goal now is real color definition, like the RGB value on Adobe Photoshop.” What inspired her effort to identify real color from DNA was a request from molecular geneticist Turi King to determine the eye and hair color of Richard III, whose remains were found under a parking lot in Leicester, England, in 2012. King, who used mitochondrial DNA to confirm the remains were Richard III’s, turned to Walsh to determine which of the portraits of the king—all painted after he was killed in battle in 1485—was the most accurate. Based on Walsh’s phenotype analysis, King determined that one of the earliest paintings of Richard III, the 1510 “Arched Framed Portrait,” best matched the genetic information.

“We were still dealing with categories [of color] because we’re not at the quantitative level yet,” Walsh said of her determination of Richard III’s hair and eye color. “[King] wanted something physical to see, and that’s what spurred me to move toward the quantitative so strongly. Because I could always say to someone, ‘blue’ or ‘blonde,’ and they would say, ‘I need to see this physically.’ So that is what I’m working on now. I want to produce that result.” Walsh has gathered DNA phenotype data from 2,000 Irish, Greek and U.S. individuals and is currently collecting data from 3,000 additional individuals from those same countries in order to create a phenotype-genotype database and prediction model. For forensic purposes, she would like to be able to start with a “blank person” and with a sample of DNA, determine the actual eye, hair and skin pigmentation.

In his Colorado lab, Spritz has been using “landmarking” to try to identify and understand the complex interactions of the genes that determine facial structure. He has used 29 facial landmarks standardized in the field as a way to correlate facial structures to genes, but he is currently moving toward a new approach. Spritz suspects that the landmark methods used in much of the facial work are not the “proper way to be thinking about facial shape because none of them have really panned out.” The previous work involving linear measurement between landmarks didn’t work. So, his new approach is to focus on landmarks that have the highest genetic component of heritability. “That is where we’ll start from,” Spritz said. “We have our fingers really tightly crossed that that will look better.”

Spritz noted that genetics is only part of what determines the appearance of a human face. “Environment and chance play big roles,” he said, although exactly how much influence they have is unclear. However, “the progressively smaller role the genes play, then the less you are going to be able to ever theoretically put everything together to make a photograph of what a person looks like.”

Shriver, who has sparred with Spritz in the journal PLOS Genetics over the feasibility of predicting a facial shape and appearance, said understanding every genetic detail, biological step and mechanism is not necessary to predict what a face will look like. “You don’t need to understand the mechanism to make predictions, to derive a statistical pattern,” Shriver said. “It is a mistake to argue that is it is too complex of a process for us to understand. The body understands it. There are all of these complex interactions, and they are read by the body. Is there something special about it that means we won’t be able to figure it out?”

Shriver noted that when humans look at one another, they are subconsciously reading genes that are the result of eons of natural selection. The face is “representative of the genetic differences between populations, and we think this is because the face played an important role in the evolution of our species,” he said.

“We interact with each other based on our faces, and natural selection determined whether we think this is an attractive face, or not. We look for faces similar to ourselves . . . and we look at the dominance of a face, whether the person looks tougher than me. These sorts of things have driven the evolution of the face at a rapid pace, so those genes have experienced accelerated evolution.”

Regardless of the focus and approach of the geneticists in their DNA phenotype research, the pursuit of the genetic underpinnings of the human face remains a daunting task. “There are some genetics that are relatively simple, like a disease,” Spritz said. “There are some that are intermediate, like a person’s height, and some that are unimaginably complex, like determining your facial shape and features.”

About the Author

Jim Dawson is a science writer with Palladian Partners, a federal contractor, on assignment at the National Institute of Justice, U.S. Department of Justice.


Date Created: September 25, 2015