Everyone once in a while I read a news report about long-lost siblings (sometimes twins) who meet for the first time as adults and discover astonishing similarities — they both love languages, they both like wearing denim skirts, and so forth. I always find those stories a bit creepy, frankly. It’s not that I deny that DNA has any effect on our personalities and tastes; but those stories often make it sound as if DNA determines absolutely everything.
So this story, traumatic as it was for the participants, cheered me up a bit:
Keeley Hall, from Perth, Western Australia, and Elizabeth Howard, from Cambridge, had been among the first to register with UK Donor Link, a Government-funded database set up in 2004.
They had been told that their DNA was one of the organisation’s first ‘matches’: they were half-sisters. Overjoyed, they told of the many resemblances between them – their similar eyes and hair, their shared love of languages – and how they already felt like sisters.
But The Mail on Sunday can now reveal that the two women are almost certainly not related at all. In a terrible and distressing mistake, UKDL brought two entirely unrelated women together and told them they were sisters.
DNA matching, the article explains, is a science of probabilities, not certainties. (I wonder how many wrongful convictions have been based on DNA?) Both of the women have since been put in contact with other women who are — really really for sure this time — their biological sisters.
Hat tip: Marriage Debate.
I having/had a similar situation happen – that is, similar in that a long-lost brother connected with us, of all things, via a picture of myself. (to note, he did quite a bit of research, and I’m odd in that my baby pictures are most definitely one-of-a-kind – and that he had one his parents gave him that I’ve never seen before – and in my culling, I’d thought I’d seen them all) and that I’d previously heard there’d been an adoptee out. He quite resembles my brother, (as in every cis sexual male from my fathers’ line is identical) but that’s where it ends on the familiarity line. He enjoys writing, as do I – particularly poetry – but that’s it. Can’t say I’ve found anything mystical between us, frankly, and he’s a sporadic emailer. (Both brothers (the new and the old one and I) decided not to tell my parents just yet – we’re waiting until my sister is out of high school and we’ve moved so it’s less hectic on this end.
I’ve got to say, while I’m glad to have made his acquaintance there’s no miraculous ‘dna’ connection, though he seems a nice enough fellow. To be entirely fair I don’t put any stock on ‘dna’ resemblence, so that has prolly colored my feelings a bit. People will find what they’re looking for – dna is, to my experience, just a handy qualifier to tack things on. Although it is nice to meet another family member who seems to give a damn.
The nice thing about probabilities is that if juries trust DNA evidence, the probabilities will tell you exactly how many wrongful convictions there have been. (Sadly, it won’t tell you which ones.) If the test claims that there’s a 99% chance that the DNA came from suspect A rather than suspect B, and you do this a thousand times, you know you’ve made ten wrongful convictions. Which is bad; but it’s not a hundred, or five hundred, wrongful convictions. So that leaves you with a value judgement: given imperfect knowledge, where do you set the threshold? You have to balance wrongful convictions because you set the standard of evidence too low against wrongful non-convictions (rapists and murderers going free) because you set the standard of evidence too high.
Also, when criticizing DNA evidence, keep in mind that at least there are estimates of its reliability; how reliable are eyewitnesses? fingerprints? footprints?
I think it’s a bit of a leap from this case, where a false inference was drawn from DNA evidence suggested these 2 individuals were 50 times more likely to be siblings than not (which still leaves a statistically significant possibility of no relationship), to a concern about false convictions based on DNA evidence. When DNA evidence is used, there are 2 questions to be asked:
1) did the person who left the DNA evidence commit the crime?; and
2) did the accused person leave the DNA evidence?
With respect to the first question, a defence lawyer can raise alternative reasons the DNA evidence was found in a particular place. But with respect to the second, assuming the sample has been properly handled, the chances of a coincidental match are very, very small. Depending on the processing method, this risk is somewhere between 1 in 5 million (early testing) and 1 in 100 billion (excluding cases where the person has an identical twin/triplets). You also need to take into account the vanishingly small probability that a person with matching DNA also would be a suspect for other reasons (ie. DNA evidence is rarely the only piece of evidence against a suspect.) So even if there are, say, 6 other people on Earth with the same DNA profile, the chances there being evidence of any of those 6 having any connection with the crime is extremely tiny.
The real risk isn’t a coincidentally matching genetic profile – it’s the possibility that the DNA got to the scene some way other than by the suspect committing the crime or that the sample was improperly handled. It’s important that proper handling procedures be maintained, and many jurisdictions have developed standard jury instructions to clarify what can and cannot be inferred from matching DNA profiles. On the other hand, DNA evidence has been extremely helpful in establishing that some wrongfully convicted people COULD NOT have committed the crime, which is why advances in DNA testing have been so important to innocence projects.
Jumping on woland’s points, I think it should also be mentioned that the DNA that is being looked at in a criminal case and the DNA in these paternity cases are fundamentally different. The former is more reliable in that the question being asked is this: is the DNA collected at the crime scene the same as the suspects? Scientists look at the sizes of dozens of polymorphisms in both sets of chromosomes (from your mother and father). They know how common each size polymorphism is in the general population and they can calculate a probability that two people share the exact same sizes. The number of sites they look at brings that range in the 1 in 100 billion.
When asking do these two women share a father, there is a lot less to go on. First, the scientists don’t know the father’s DNA. And each prospective sibling could have gotten any combination of chromosomes from their father (who has two, one from his father and one from his mother). Also any one of those chromosomes could have undergone recombination, further complicating the issue. Lastly, it is conceivable that they could share no DNA and still be half sisters by the freakish possibility that one sister only got DNA from her father’s mother and the other only got DNA from her father’s father (well, they’d have to share the X but you get the idea).
So the unreliability of DNA in criminal cases really is in improper handling of the samples and not in the probabilities.