DNA contains all sorts of ‘intelligence’ useful to criminal investigators. Most traditional analysis has focused on DNA profiling, also called DNA fingerprinting; this relies on repeat sequences of DNA – called short tandem repeats, or STRs – which, like a fingerprint, are so varied that they identify an individual. Once the police have a suspect’s DNA, they can confirm whether there is a match between their STR profile and the STR profile in a biological sample found at a crime scene – usually miniscule amounts of blood or semen, saliva or hair.
However, DNA fingerprinting tells police nothing about what the person looks like. Now it seems that new tools and technologies are helping investigators to do just that - to go beyond DNA profiling to predicting actual physical appearance.
Single nucleotide polymorphisms (SNPs) are regions of DNA where there is variation at a single nucleotide (A, G, C or T) site between people. This can affect physical characteristics and a suite of SNPs differences can be used to make predictions about physical appearance.
In 2011, the lab of forensic geneticist Manfred Kayser at Erasmus University in Rotterdam, the Netherlands, developed a test called IrisPlex that accurately predicts for blue or brown eye colour, based on just six SNPs (Forensic Sci. Int. Genet., doi: 10.1016/j.fsigen.2010.02.004).
People had brown eyes until Europeans began dropping the melanin pigment and evolved blue eye colour, perhaps 10,000 years ago. ‘We, and other groups, figured out that there is a SNP located next to one of the main pigmentation genes that was the most correlated with eye colour,’ explains Susan Walsh, a former student in Kayser’s lab and now professor of forensics at Indiana University-Purdue University, Indianapolis, US. ‘It basically acted as a regulatory switch turning it from brown to blue.’
In 2013, the group went a step further and IrisPlex was developed to simultaneously predict eye and hair colour, based on 24 SNPs (Forensic Sci Int Genet. 2013, 7, 98). Eye colour can be predicted at around 95% accuracy for blue, 75% for brown, while black and red hair can be predicted at 90% accuracy and blond or brown 75-80%. Walsh is now working on skin pigmentation and tests for discriminating shades of hair colour.
Height is another characteristic determined by individual SNPs. Height is 80% genetically inherited, and 700 SNPs have been identified that together explain 16% of variation in height. In forensics, large numbers of SNPs can be impossible to obtain; specks of degraded blood may be all the DNA evidence at a crime scene and getting good quality DNA is often a problem.
‘The forensic field is interested in minimising the set of SNPs,’ Kayser says, pointing out that ancestry detection is now possible with small sets of 20 or 30 SNPs. It is possible to determine bio-geographic ancestry of a person at a continental resolution, telling Europeans from Oceania from Africans.
But SNPs are not the only way of interrogating DNA. Scientists are increasingly interested in epigenetics: nature’s way of tuning gene activity by adding methyl groups to specific places on DNA strands. Proteins bind to the methylated DNA, causing a physical change to DNA that turns genes off. For forensic scientists, patterns of DNA methylation could reveal the activities of telltale genes and give clues to pathological states, circumstances leading to death or tissue type (A. Vidaki, Forensic Sci. Intern.: Genetics, 2013, 7, 499). Various diseases and cancers are also associated with abnormal methylation patterns (M. Szyf, J. Neurodev. Disord., doi: 10.1007/s11689-011-9079-2), as are neurological disorders, which are a symptom of lead poisoning (D. Fragou et al, Toxicol. Mech. and Methods, doi:10.3109/15376516.2011.557878).
Methylation patterns may also help investigators determine body fluids (J. L. Park et al, Forensic Sci. Internat.: Genetics, doi:10.1016/j.fsigen.2014.07.011). This is important in sexual assault and rape cases where it can be important to say what cells or body fluids are present when a mixed DNA profile – often both the victim and assailant – is obtained. A swab from a victim may contain skin, semen, blood, hair, epithelial and/or saliva; knowing what you have in a crime scene sample can be crucial.
In 2014, scientists in China went a step further by reporting the building of a model they say explains 95% of variance in age. The model is claimed to estimate a person’s age, plus or minus five years, from analysis of just three DNA methylation sites (S.H. Yi et al, Int. J. Legal Med., doi: 10.1007/s00414-014-1100-3).
‘If you use 50 to 100 sites, accuracy goes up to plus or minus two or three years, but that involves a lot of material and is not realistic, unless you had a lake of blood, and that is typically not the case,’ says Peter Schneider at the Institute of Legal Medicine at the University of Cologne, Germany. Analysis of DNA methylation requires a lot of DNA, he explains. ‘You need to do a bisulphide conversion, to convert the methylated bases into a form you can detect through DNA sequencing.’
Another means of estimating age, meanwhile, proposed by Kayser’s group, involves measuring the decline in certain circular DNA pieces in defensive T cells (Current Biology, doi: 10.1016/j.cub.2010.10.022).
Arguably the most exciting new forensic development, however, is the potential to place a photo into the hands of investigators via a DNA test. In the future, ‘you won’t need DNA databases anymore,’ says Kayser. ‘You will get an image and run it against the passport pictures, with names and addresses, and the next day the police can go knock on the door.’
Right now, there isn’t much knowledge about genes that underlie appearance, but scientists like Kayser and Walsh are beginning investigations. ‘We look at landmarks, specific measurements of faces,’ says Walsh. ‘So we see how individuals differ from each other and try to find the genes related to those landmarks.’
A 2014 paper by Mark Shriver and colleagues at Penn State University, US, analysed 76 genes linked to facial abnormalities and identified 24 markers for use in a 3D modelling program to help predict facial shape (PLoS Genetics, doi: 10.1371/journal.pgen.1004224). However, the study attracted criticism for basing predictions on markers linked to ancestry, which aren’t reliable to predict how someone looks. ‘The results were overstated and I question whether they even looked at face genes,’ says Kayser.
The company Parabon NanoLabs in Virginia, US, is offering to generate investigative leads by predicting physical appearance from a DNA sample, based partly on the Shriver research. In January 2015, Columbia, South Carolina, became the first police department to release a face prediction publicly, for a murder case of a mother and three-year-old daughter. However, experts such as Kayser dismiss Parabon’s prediction as unreliable, and worry the claims will create false expectations and set the technique back just as research begins in earnest.
‘Ancestry should never do phenotyping, and phenotyping should never do ancestry. They should be used on their own,’ says Walsh. Phenotype is the physical outcome of our genes, or genotype. Facial differences will be explained by a large number of DNA variants, each with a small effect, similar to body height, according to a genome-wide study reported in 2012 by a European consortium of scientists interested in visible trait genetics (F. Liu et al, PLOS Genetics, doi: 10.1371/journal.pgen.1002932).
The Netherlands introduced regulations allowing DNA ancestry information and physical appearance predictions to assist investigations as far back as 2003.
It was first used when DNA testing indicated the murderer of a 16-year-old Dutch girl was of western European descent, a finding that challenged town residents’ suspicions that the perpetrator came from a nearby hostel for non-European asylum seekers.
Dutch police today can ask their forensics labs to run hair and eye colour and biogeographical ancestry predictions. This is not the case elsewhere.
‘In Germany, you can only look at DNA markers [for matching two profiles] and should not explore any information beyond that,’ says Schneider. ‘Holland has got this right, but Germany hasn’t. Our lawmakers do not want to touch this issue. Some of this is connected to Nazi history, when people were discriminated against based on so-called race.’
‘The legal issue will be the main barrier in many countries,’ Schneider predicts. ‘Look five years ahead and pigmentation markers and ancestry prediction will be commonplace in countries which allow it. Facial prediction will need another five years after that, but then it will become really interesting.’