Posted Monday, March 30, 2009, at 7:50 AM
Is it OK to guess a perpetrator's race and appearance from DNA found at the crime scene?
A few weeks ago, Gautam Naik of the Wall Street Journal updated us on a fertility clinic's program to screen embryos for " eye color, hair color and complexion ." The clinic hoped to use a method that could supposedly "identify [genes] that relate to northern European skin, hair and eye pigmentation in 80%" of IVF embryos.
Two weeks later, when the clinic suspended the program, I suspected its concession was more technical than moral. Predicting traits from genes is hard. That's one reason why so many threats and promises of genetic engineering haven't materialized.
But now it seems I may have underestimated the field. Naik has returned with further research on the genetics of appearance. Gene-trait correlations are becoming increasingly precise. And the practice to which they're being applied most aggressively isn't embryo screening. It's law enforcement.
Here's his latest report :
Murray Brilliant, a geneticist at the University of Arizona, is developing a predictive test for skin, eye and hair color. Supported by a grant from the U.S. Department of Justice, he and his colleagues recently analyzed DNA material provided by 1,000 university students from different ethnic backgrounds. They found a total of five genes that account for 76% of the variation for hair color, 75% for eye color and 46% for skin color. Similarly, scientists from Erasmus University published a paper in March in the journal Current Biology based on a DNA analysis of 6,000 people in the Netherlands. For that population group, they found that only six DNA markers are needed to predict brown eye color with 93% accuracy and blue eye color with 91% accuracy.
Some of those numbers are quite impressive. If they're packaged into an affordable embryo test, I guarantee you buyers will start lining up. But look where the money's coming from. Brilliant got his grant from DoJ. The Erasmus group got its funding from the Netherlands Forensic Institute , which "provides services to clients within the criminal justice chain, such as the Public Prosecution Service and the police ." In fact, Naik points out, "[t]he push to predict physical features from genetic material," known as "DNA forensic phenotyping," has
already helped crack some difficult investigations. In 2004, police caught a Louisiana serial killer who eyewitnesses had suggested was white, but whose crime-scene DNA suggested—correctly—that he was black. Britain's forensic service uses a similar "ethnic inference" test to trace murderers and rapists. In 2007, a DNA test based on 34 genetic biomarkers ... indicated that one of the suspects associated with the Madrid bombings was of North African origin.
Whoa, there. Do we really want cops hunting for people of a particular race or ethnicity based on uncertain inferences from a DNA sample? Apparently several states don't. Nor does Germany. These jurisdictions, according to Naik, forbid "the forensic use of DNA to infer ethnicity or physical traits."
I understand the concern. But these prohibitions are a mistake. DNA forensic phenotyping doesn't need to be perfect. It just needs to be better than—or a substantial way of double-checking—the unscientific inferences we already make.
The Louisiana case sets off racism alarms because DNA phenotyping said the culprit was black, whereas witnesses said he was white. But that isn't the usual pattern. Decades of research suggest that a witness is more likely to misidentify a person as the perpetrator when the accused person is of a different race . And according to the Innocence Project ,
Racism continues to be a significant cause of wrongful convictions. While 29% of people in prison for rape are black, 64% of the people who were wrongfully convicted of rape (and then exonerated through DNA) are black. Moreover, most sexual assaults nationwide are among perpetrators and victims of the same race (the federal government says just 12% of sexual assaults are cross racial), but two-thirds of all black men exonerated through DNA evidence were wrongfully convicted of raping white people.
DNA phenotyping didn't invent the problem of erroneous racial inferences in law enforcement. That problem is as old as racism and lives on through mug books, lineups, and eyewitness testimony. As the Innocence Project points out, DNA is beginning to clean up the problem. Granted, DNA phenotyping isn't nearly as reliable as DNA matching. But is it really worse than relying on witnesses alone? At the very least, wouldn't it be useful and wise to check witness recollections of the perpetrator's race or ethnicity against a DNA phenotype analysis?
The same goes for facial features. Naik reports:
Mark Shriver, an anthropologist and geneticist ... [is] trying to construct a "picture" of a person's face by analyzing DNA. He calls the technique "forensic molecular photo fitting," and it is supported by a $500,000 grant from the U.S. Department of Justice. ... His team collected DNA samples and photographs from 243 people ... and used computer techniques to correlate the genes with his subjects' facial features. They have found six genes that seem to influence such traits. One gene is associated with the height of the face; another is associated with its width. Yet another gene affects the shape of the lips and the nose. By piecing together these elements, Prof. Shriver hopes to create a modern-day version of the police artist sketch.
Initial results of this method will probably be pretty crude. But will it end up being worse than old-fashioned police sketches based on eyewitness accounts? Would we have caught the Son of Sam killer earlier, for instance, if his police sketch hadn't been wildly inaccurate, making him look Latino or Asian ?
No, DNA phenotyping isn't perfect. But it's better than nothing. And it's better than trusting witnesses alone.