DNA - a silent witness

Forensic science is about to be transformed with the use of new technology to predict a person’s physical appearance - a molecular photofit derived from DNA found at a crime scene.

On New Year’s Day, 1999, a 92-year-old woman was bashed and raped in her home in the small NSW town of Wee Waa, population 1650.

Police suspected the rapist was a local and narrowed the search to a dozen suspects but could go no further. Finally, in April 2000, after 14 months of inconclusive investigation, they took the unprecedented step of asking all local males between the ages of 18 and 45 to voluntarily provide a saliva sample for DNA testing.

Ten days later, before the DNA analyses were complete, one of the donors - a 44-year-old farm labourer and father of three, Stephen James Boney - went into the Wee Waa police station and confessed, after realising he was certain to be found out.

The cost and logistical challenge involved in testing all males of a certain age makes such an approach impractical in large communities.

But what if, instead of interrogating a list of suspects, police could profile an anonymous criminal’s DNA, found at the scene, and within hours, produce a generalised description of the unknown culprit: their hair and eye colour, the length of their nose, the distance between their eyes and even their ancestry?

A molecular photofit

University of Canberra forensic geneticist Dr Dennis McNevin is leading a three-year collaborative research project to develop one of Australia’s first DNA-based ‘photofit’ systems, based on single-nucleotide polymorphisms - the ultimate source of human individuality.

The Australian Research Council-funded project involves researchers from the University of Canberra, the Queensland Institute of Medical Research, the Australian Federal Police and the Victoria Police. McNevin says both police forces are keen to apply the new technology when the project delivers the first working version of the genotyping platform in around three years’ time.

Many crimes go unwitnessed. McNevin says even if there are witnesses, their descriptions of the culprits are notoriously unreliable.

Often, the only clue to the identity of the perpetrator may be traces of DNA from the scene. Standard DNA ‘fingerprints’, based on microsatellite DNA sequences scattered across the genome, can only identify a suspect if his or her DNA sample matches a signature held in a database of DNA signatures of convicted criminals. If the unidentified suspect has no criminal record, the DNA trail ends there.

But an anonymous DNA trace from a crime scene contains phenotypic information about a suspect - McNevin describes it as a ‘silent witness’ that can be interrogated to produce a generalised description of the perpetrator.

He and his colleagues hope to provide police with the tools to assemble a DNA-based photofit of a suspect who has left a trace of their DNA at the crime scene.

McNevin says that only in very restricted circumstances would a generalised DNA photofit directly identify the perpetrator of a crime.

“What we hope to do, a long way down the track, is identify a few hundred to perhaps one thousand SNPs from which we could put together an image of someone’s face, using physiological metrics,” he said.

Shades of blue

The panel of SNPs will be used to produce primers for a PCR assay kit to detect corresponding SNPs in DNA traces recovered from the crime scene; special software will transform the pattern of hits into a DNA photofit of a suspect.

Various SNPs involved in melanogenesis (the genetic pathway that synthesises melanin) have a major influence on eye, hair and skin colour, although only the first two characters are useful for a DNA photofit because of the range of variance in skin hue with exposure to ultraviolet radiation.

Eye and hair colour also vary, and McNevin says the regression technique used to predict these traits from SNP variants will provide a categorical rather than a definitive prediction - for example, if the subject’s SNP profile confirms he or she has blue eyes, it may only be possible to categorise the hue as a pale or darker shade of blue.

“A regression is only as good as the data set, and the bigger the data set, the better,” he said. “The important thing is to devise a rigorous classification system so that when the computer makes a prediction, it can be translated into a very objective way of representing the subject’s phenotype.

“We don’t care what the genes are - we’re interested only in the SNPs associated with each particular phenotype.”

Matching facial metrics

McNevin says magnetic resonance-imaging (MRI) scans of the size and shape of human skulls have led to the identification of a number of SNPs that strongly influence facial metrics like the distance between the eyes and the length of the nose.

The QIMR’s main contribution to the project thus far has been the discovery of a set of SNPs found in seven genes associated with male pattern baldness.

The QMIR research project has shown that X does not mark the bald spot - several of the genes are located on autosomes, scotching the popular misconception that maternally inherited genes on the X-chromosome are solely responsible for male susceptibility to pattern baldness.

“In an ideal situation, we might be able to collect DNA from a crime scene, genotype it for key SNPs and upload the resulting photofit to airports and border crossings,” McNevin says.

“Video cameras linked to face-scanning software could identify individuals whose facial metrics are a near match to the photofit and they would be taken aside and interviewed by police - that would be a great boon for forensic science.

McNevin says an SNP-based photofit can also be very useful in the inverse situation, where police need to narrow a large field of candidate suspects, by excluding those who do not match the basic set of physical characteristics inferred from a DNA sample.

He says the approach his group has adopted means that as more data are added to the original training set of SNPs, the phenotype predictions can be refined, leading to more accurate photofits.

Relatively few phenotypic traits are monogenic; most emerge from the concerted action of multiple genes.

Given that phenotypes are also strongly influenced by environmental factors, McNevin says it is effectively impossible to predict how multiple alleles of multiple genes interact with environmental factors to produce highly variable phenotypic features like body weight.

Even a strongly heritable trait like a person’s height is influenced by multiple genes - although the dominant influence of one or two genes like the human growth hormone (HGH) and insulin-like growth factor (IGF-1) genes may permit a ballpark estimate of a suspect’s stature.

Tracing ancestry through mitochondrial DNA

In a paper published in the Australian Journal of Forensic Science (AJFS) in March 2011, McNevin and several colleagues from the University of Canberra, the Australian Federal Police and Victoria Police describe how genetic markers in maternally inherited, non-recombining mitochondrial DNA (mtDNA) can provide a detailed record of the ancestry of females and males.

McNevin says ancestry can be greatly refined with the additional use of the non-recombining part of the Y chromosome (NRY) as well as autosomal markers.

The most celebrated project to use SNP markers from mtDNA and NRY DNA is the National Geographic Society’s Genographics Project. It has assembled a detailed record of the epic migrations of our ancestors as they moved out of Africa to colonise Europe, Asia, the Americas, Australia and the Pacific over the past ~150,000 years.

In the absence of recombination, mtDNA is transmitted down the female line as a stable haplotype, subject to occasional, random single-nucleotide mutations.

Over tens of thousands of years, mutations accumulate that, because of the genetic founder effect, effectively constitute a record of when and where an individual’s direct female ancestors lived.

The earliest non-African mtDNA mutations reveal the deep ancestry and early dispersal of female ancestors from their locus of origin.

Over time, migration and genetic admixture have caused early lineages to intersect geographically, to an extent that no mtDNA haplogroup is exclusively present even in the most isolated regions - although haplotype B predominates in Polynesians, who emerged from an ancestral homeland in Taiwan 5000 years ago to colonise the Pacific.

Sharing common ancestors

Lineages inferred from mtDNA and NRY haplogroups typically tell different tales. McNevin et al say historical differences in male and female migration patterns mean that mtDNA lineages are not as well correlated geographically as NRY lineages, because of patrilocality - the tendency of males to remain within the locality of their male ancestors, while females disperse.

They note that ancestry profiling by mtDNA SNP genotyping is not definitive - correlating haplogroups with genealogy requires surveys to determine population-defining SNP patterns.

Researchers have already defined the major European haplogroups for individuals from populations in Austria, Spain, Italy and the USA, and the East Asian haplogroups for individuals from Japan, Korea, China, Taiwan, and Asian immigrants in Argentina.

In comparison, haplogroups for Australia and Oceania have been poorly documented. The AJFS paper outlines the results of their SNP genotyping survey of 145 individuals, who voluntarily provided details of their female ancestry.

The study yielded a multiplex SNP assay that differentiated between Asian, Caucasian and African populations within Australia. Haplogroups M, B and F predominate in individuals of self-declared Asian ancestry, haplogroup L predominates in those of African ancestry, whereas haplogroups H, T and U predominate in non-Indigenous Australians of European ancestry.

“The endpoint of the project is that we would like to be able to settle on a stable, high-throughput genotyping technology for use with the available data - we’ll be happy if we can accurately predict ancestry and physical characteristics such eye and hair colour, detached or attached ear lobes, and a cleft chin.

“But in five years’ time, as the data become available, we might be able to include many other phenotypes.”

*****************************************************

SNP PROFILING - a case history

The first successful application of SNP profiling to help identify the perpetrator of a serious crime was in Baton Rouge, Louisiana, in 2002. As much as it helped police to identify a culprit, it also demonstrated the technique’s power to eliminate false leads.

On 17 July, a group of workers walking along the Mississippi River west of Baton Rouge found the naked body of a young woman. Pam Kinamore had been abducted from her home in Baton Rouge three days earlier.

Traces of semen recovered from her body revealed she was the third victim of an unidentified serial rapist-murderer operating in the area, but investigators were unable match the murderer’s microsatellite DNA profile to any profile in the CODIS (Combined DNA Index System), a national database containing the coded microsatellite profiles of 5 million known criminals.

After Kinamore’s body was found, a female witness reported seeing a young white man driving a white pickup truck in the vicinity around the time of her disappearance, with what appeared to be the body of a woman slumped in the passenger seat. Police regarded it as a strong lead, because another woman had been picked up and raped by a man driving a white pickup truck two days after Kinamore was abducted.

Despite the lead, police had no prime suspect when the murderer struck again in December.

Investigators obtained DNA samples from 1200 young white men from the area. But months of analysis at a cost of more than US$1 million failed to produce a match to the murderer’s DNA.

On the increasing probability that coincidence and a flawed witness report had misled police investigating Kinamore’s murder, Dr Tony Frudakis, founder of a struggling biotechnology company, told police he could determine the killer’s race from a DNA sample, with 99% certainty.

Detectives tested Frudakis’s claim with a double-blind test, sending him coded samples of DNA from 20 individuals of Caucasian or Afro-American ancestry. Using a set of 175 SNPS from loci known to be informative about an individual’s racial background, Frudakis aced the test.

Frudakis then tested the killer’s DNA and informed police the killer was of Afro-Caribbean or Afro-American descent - not Caucasian.

By then, the rapist had killed again. After almost a year pursuing a false lead, police switched focus to an imprisoned sex offender, previously excluded by his Afro-American ancestry.

Four days before Pam Kinamore was abducted, Derek Todd Lee, 34, had been interrupted as he attempted to rape another woman in her home; he fled, but the woman was able to provide an accurate description of her assailant.

Despite the almost identical circumstances to Kinamore’s rape and murder, police did not attempt to obtain a sample of her assailant’s DNA, much less test it for a match with the DNA profile of Kinamore’s killer, whom they believed to be white.

When he was eventually arrested, Lee was carrying a gun. He was sentenced to prison for two years. Police subpoenaed Lee in prison for a cheek swab, and tests confirmed his DNA profile matched that of Kinamore’s killer.

*****************************************************