Creating a Face from a DNA Sample

Create Face banner 1

Posted: June 12, 2017 | By: Michael Leuze, Ryan Tokola

The Department of Defense (DoD) continually seeks new and emerging methods to improve biometrics for enhanced security and identification. Biometric facial recognition technology analyzes collected information and establishes unique facial characteristics to compare against watch lists of suspected or known terrorists for identification purposes.

Some existing technologies use DNA to identify demographic characteristics, such as gender and ethnic background, and create a face prediction consistent with these general attributes. This approach is of limited use for identification as the predicted faces do not account for the individual’s specific genetic information.

In contrast, the ongoing research at the Department of Energy’s Oak Ridge National Laboratory (ORNL) focuses on interpreting portions of DNA that govern the unique facial structure of each person.

Traditional facial recognition biometrics techniques use cameras or sensors to capture an image before comparing it to a previously established database of known suspects in order to identify unknown persons. [1] Likewise, DNA samples are typically useful in biometrics only when refBy: Ryan Tokola & Michael Leuze, Ph.D. 10 B erenced back to a pre-existing database of identified DNA.

Researchers at ORNL are moving the idea of creating a face from a DNA sample closer to reality with a novel method that assesses facial scans and compares them to genetic markers.

The ability to create a facial image based solely on small amounts of genetic material, without the need for pre-existing images of the individual or extra equipment, provides a huge advancement in current biometric capabilities.

ORNL’s DNA2Face project demonstrates that predicting a person’s facial structure from their genetic data, such as a small blood or tissue sample, is possible. The advanced identification possibilities brought forth by this technology will benefit the DoD community as well as law enforcement, national security and the intelligence community.

DNA2Face Project The goal of the initial DNA2Face project was to determine whether a significant association exists between DNA and unique facial features. In order to demonstrate the association, ORNL researchers created a statistical method to represent the 3D shape of faces.

The researchers’ novel biometric approach first defined a reference, average face, mapped with tens of thousands of 3D points. The reference face was then aligned with a similarly mapped facial scan. Comparing the two created correspondence vectors (CV) — essentially a measure of the differences between the reference face and the scanned face.

Once the CVs were established, the researchers applied the statistical technique, principal component analysis (PCA), which transformed each CV into a relatively small set of numbers, providing a very compact representation of the face.

The researchers then performed a genome-wide association study (GWAS) using DNA samples obtained from the subjects of facial scans. A GWAS is a statistical approach to finding genomic variations that are
associated with a particular trait. The results of the GWAS showed several strong connections between DNA and facial features by comparing PCA scores to DNA markers. [2] The researchers found 30 mutation locations with genome-wide significance and more than 5,000 locations with suggestive significance. In comparison, an earlier study without ORNL’s statistical method found only five regions with genome-wide significance. [3]

“These days if you get a genetic sample from a crime scene, basically all you can do is apply it to some databank of samples, and we don’t have a lot of those. But if you can take that genetic material and make a face prediction then you can run it against a database of faces, which we have a lot more of,” notes Principal Investigator Ryan Tokola of ORNL’s Imaging, Signals and Machine Learning Group.

Recognition and Prediction

Using facial recognition technologies is not a new concept for the DoD. From1993 to 1997, the DoD’s  Face Recognition Technology (FERET) program’s primary mission was to develop automatic face recognition capabilities able to assist security, intelligence and law enforcement personnel in performing their duties. [4]

The ability to identify an individual from genetic material would enable the DoD to
identify extremists from remaining touch DNA or any genetic material left behind. Furthermore, the ability to create a facial image without a database of known persons could allow the military to track and identify extremists operating near deployed forces. With the rise of lone-wolf terrorism, it is impossible to keep a comprehensive record of individuals who may commit acts of terrorism. The ability to identify suspects with only genetic material available will assist officials in tracking and apprehending persons of interest.

The DNA2Face method could also be deployed to identify unknown remains of service members killed in action. Additionally, during special operation missions, the DNA2Face technology could provide target confirmation after military operations.

The Details

As technology reduces the time and cost required to sequence DNA and identify genotypes, researchers are better able to produce massive databases containing human genomics information. This allows portions of genetic code to be correlated to specific physical traits, or phenotypes.

Utilizing vectors to define facial characteristics allows efficient comparison of genetic variations from different individuals with a variety of traits. Employing a GWAS allows for the rapid scanning of the genome to pinpoint markers and genetic variations or the same feature, such as facial shape.

The GWAS array used in the study consisted of more than 700,000 single-nucleotide polymorphisms, or small variations in the genome. [5]

In order to minimize cost and to analyze many faces to identify correlations between phenotype and genotype, researchers used pre-existing data sets via the FaceBase Consortium. [6] Phenotype datasets provided 3D face meshes in a Wavefront format. Each of the faces were oriented facing the –y direction with the origin of the coordinate system located either behind or in front of the facial plane. “Because the coordinate points are very close together, and there are tens of thousands of them, it results in a very dense representation of the face,” Tokola explained. Demographic characteristics such as ethnicity, gender and age were taken into account for the subjects included in the research. The most significant cluster of variations was found on chromosome 3 on the RAF1 gene, which is known to affect craniofacial shape. Researchers suggest analyzing select portions of the face could result in the discovery of more genes that directly influence facial shape.

Conclusion

Researchers at ORNL have demonstrated a strong association between genetic data and facial morphology. Going forward, a much larger dataset will be required to account for genetic variations in order to accurately predict facial shape and other finer features. This predictive capability would be a significant step forward in the application of biometrics to several security and intelligence-related problems. With additional research and technological advances such as high-powered computing, the ability to map a complete face with genetic material is possible.

The ability to recreate faces based on DNA sequences would revolutionize the way unknown individuals are tracked and monitored. This technology could also benefitborder and airport security.
Additionally, this technology could be used by the DoD to confirm and identify extremists and positively identify service members killed in action.

1. Woodward, J. D., Jr., Horn, C., Gatune, J., & Thomas, A. (2003). Biometrics: A look at facial recognition. Santa Monica, CA: RAND Corporation. Retrieved from http://www.rand.org/pubs/documented_briefings/DB396.html (accessed May 1, 2017).

2. National Human Genome Research Institute. (n.d.). Genome-wide association studies. Retrieved from https://www.genome.gov/20019523/ (accessed May 1, 2017).

3. Liu, F., Lijn, F. V., Schurmann, C., Zhu, G., Chakravarty, M. M., Hysi, P. G., . . . Kayser, M. (2012). A genome-wide association study identifies five loci influencing facial morphology in Europeans. PLOS Genetics, 8(9). doi:10.1371/journal.pgen.1002932

4. National Institute of Standards and Technology. (n.d.). The Facial Recognition Technology (FERET) database. Retrieved from http://www.itl.nist.gov/iad/humanid/feret/feret_master.html (accessed May 8, 2017).

5. Genetics Home Reference. (n.d.). What are genome-wide association studies? Retrieved from https://ghr.nlm.nih.gov/primer/genomicresearch/gwastudies (accessed May 1, 2017).

6. Boehnen, C., Peters, T., & Flynn, P. J. (2009). 3D signatures for fast 3D face recognition. In M. Tistarelli & M. Nixon (Eds.), Advances in Biometrics. ICB 2009. Lecture Notes in Computer Science, 5558, 12-21. Springer Berlin Heidelberg. Retrieved from https://link.springer.com/chapter/10.1007/978-3-642-01793-3_2 (accessed May 1, 2017).

Focus Areas