A study published on March 20, 2014 reports on the possibility of computers to accurately create virtual mugshots of an individual based solely on that person’s DNA. Several scientists from around the world, the majority of them from Belgium and the U.S.A., have studied facial features based on the aspects of ancestry, gender and individual genetic traits to come up with a formula for accurately predicting facial reconstruction from saliva, skin or a strand of hair. This means that if this initial study is improved upon and the process refined, in the future (one author of the study claims this may happen within the next 5 to 10 years), accurate criminal mugshots may be created for police and other investigators from genetic material found at the scene of the crime.
A virtual mesh was created of a generic human face; this mesh contained equally-sized, fine points of space that scientists refer to as “quasi-landmarks.” Essentially, each quasi-landmark is a single point of space within the mesh, which is made up of 7,150 of these points of space. The result looks much like a delicate webbing of chain mail.
Researchers took 600 volunteers from the United States, Cape Verde and Brazil. They more specifically chose people from these geographic regions who were of mixed European and West African ancestry. A stereoscopic camera took 3D images of each volunteer. Colour and fine details such as eyebrows and other facial hair were not included in the end result, creating a sort of virtual plaster-of-Paris bust of real, individual faces.
To ensure that the end result is symmetrical and not off balance, scientists created a mirror-image of the initial 7,150-point mesh. The original and mirror-image are both used when a 3-D version of the mesh is created. The 3D version ends up therefore containing three times the number of points, equaling 21,450 separate points of space. This was placed virtually over each of the plaster-of-Paris looking busts. The 21,450 points on each face acts as a coordinate system, in the way that intersecting points of latitude and longitude are used to create GPS coordinates.
The fairly blank images of each of the 600 volunteers was presented, one by one, to another set of volunteers who rated facial structure (including specific facial regions) according to measures of perceived ethnicity and gender. This is because scientists already know many of the genetic markers linked to ancestry and gender, and so could then isolate results not caused by either of these. This enabled them to come up with a way to extract the effects of individual genes known to be related to facial structure that are also known not to be related to either gender or race.
Scientists found that up to one third of the variance in facial aesthetic can be explained by issues of ancestry and gender. This leaves a large amount of variance to be ascribed. Unfortunately, one has to also consider factors of age, environment and weight when considering the end result of any individual face. The test cannot, of course, do this, which necessarily limits the creation of the perfect algorithm to predict accuracy close to 100%. Nonetheless researchers were able, based on: (i) their model for the measurement of facial coordinates, (ii) volunteer reporting and (iii) pre-existing knowledge of genes related to facial structure, to create an algorithm for the approximation of facial structure. As race and sex are known from DNA (as are hair and eye color), these are entered into the algorithm from the genetic material. The formula then fills in the blanks, so to speak. The end result is thus far still crude, but is a fair representation of what a person with specific DNA might look like.
On September 4, 2013, CNN reported on a New York artist named Heather Dewey-Hagborg who was already doing exactly what this report describes. Likely simultaneously to the development of the technology being studied in the report, Dewey-Hagborg took her own DNA, and subsequently that of other people found in public places, to a biotechnology lab where 3D faces were created from the genetic material. Once sequencing is done, the printing process within the lab takes eight hours. The study published on March 20 was submitted (to the medical journal it ended up being published in) eight days after the CNN article on Dewey-Hagborg’s project.
More research is necessary to finesse things. Specifically, a greater number of individuals across as many races as possible need to be studied in a similar way. In time and with further studying, scientists may be able to more specifically identify which particular genes in each individual are responsible for curvature of the nose (many genes may be involved in one aspect of a feature), jaw shape, and so on. Scientists are hopeful that fairly accurate virtual mugshots will be created from DNA in the not-too-far-off future.
By Julie Mahfood
Follow Julie Mahfood on Twitter @JulieWrites2