Are DNA ‘cold hits’ resulting in miscarriages?
From CrimeLine from Andrew Keogh
Summary:
The UK National DNA Database is now so large that false DNA matches are actually predicted. The number of false matches predicted is low, only two or three, and the official position is that no false matches have occurred. However, this prediction is likely to be a significant underestimate and I am concerned that false DNA matches will lead to miscarriages of justice, if they have not done so already.
The UK National DNA Database (NDNAD) is the largest in the world. As at 30 September 2008 the estimated number of individuals with records (DNA “Subject Samples”) on the NDNAD stood at a staggering 4,343,624, and by 31 March 2009 this number is likely to have approached or exceeded 5 million. These Subject Samples represent individuals who have either, had DNA samples taken when arrested or charged with recordable offences, provided DNA as volunteers and consented to their DNA being retained on the NDNAD, or come from prisoners who have had their DNA taken as part of the serving prisoner DNA sampling program. According to the 2006-2007 NDNAD Annual Report the number of DNA profiles from crime scenes (DNA “Crime Scene Samples”) held on the NDNAD stood at 285,848 as at 31 March 2007. However, crime scene samples are removed from the database when the crime is ‘solved’. The total number of Crime Scene samples uploaded as at 31 March 2007 was 427,437. Given the past rate of crime scene sampling, it is likely that the number of Crime Scene samples actually uploaded will easily have exceed 500,000 by 31 March 2009.
Whenever a new Subject Sample is loaded onto the NDNAD it is compared with all the existing Crime Scene Samples to cross-check for a match. Correspondingly, whenever a new Crime Scene Sample is loaded onto the NDNAD it is compared to all the Subject Samples to cross-check for a match. Crime Scene Samples are also compared to all the other Crime Scene Samples to check if crime scenes are linked. The number of comparisons being made as samples are continually uploaded to the NDNAD is enormous. For example, each time a new Crime Scene Sample is loaded it is automatically compared to the five million odd Subject Samples. The next crime scene sample is also compared to the five million Subject Samples. Assuming the two crime scene samples are different, then 10 million comparisons have been made from the uploading of just two fresh Crime Scene Samples.
Due to the current size of the NDNAD, full profile chance matches (“adventitious matches”) are now actually expected to have occurred. A truly staggering 2.5 trillion odd comparisons have been made: 500,000 (Crime Scene Samples) x 5,000,000 (Subject Samples) = 2,500,000,000,000. Under ideal conditions, a full SGM+ profile1 is generally predicted to occur with a frequency of 1 in 1 trillion (although a figure of 1 in 1 billion is routinely given in court as it is perceived to be highly conservative in favour of the defence). With the NDNAD at its current size, and taking the 1 in 1 trillion chance match figure, a simple mathematical analysis indicates that two full SGM+ matches are expected to have occurred between a Crime Scene Sample and an unconnected Subject Sample. Put another way, it is expected that two suspects have been wrongly identified as linked to a crime scene by a false match.
Unfortunately, a simple mathematical analysis will not give the true picture. Two factors will increase the probability of adventitious matches: firstly, the condition of crime scene samples may lead to incomplete profiles; and secondly, individuals who are related are more likely to share the same profile than unrelated individuals.
A crime scene sample can be uploaded onto the NDNAD where results have been generated at only eight out of 10 loci (loci are DNA analysis sites), and a speculative search can be conducted where only six of 10 loci have produced results. Where results are obtained from less than the full 10 loci, the random match probability is reduced; it is more likely that a match will occur by chance alone. The comparison of incomplete profiles suggests that the prediction of only two adventitious matches could be a significant underestimate. Depending on the number of incomplete Crime Scene Samples uploaded, the number of expected adventitious matches could easily run into double digits. It would also appear that the problem of adventitious matches is set to increase with the expansion of low copy or low template DNA tests (“LtDNA”). LtDNA testing frequently produces incomplete profiles. If these incomplete profiles are uploaded to the NDNAD, or speculatively searched against it, the probability of obtaining adventitious matches will increase. Even more concerning is the fact that in a LtDNA test the entire DNA sample collected from a crime scene can be consumed, preventing further testing that would exonerate a person wrongly linked by an adventitious match.
People who are related share genetic material derived from their common ancestors. The closer the relationship the greater the chance their DNA profiles will match. In court, this would be expressed by a reduced match probability for a scenario that a relative committed the crime (e.g. match probability 1 in 1 billion for an unrelated individual, but one in several hundred thousand for a full sibling). It is a known fact that the NDNAD contains the profiles of a large number of related individuals. However, due to the level of duplication on the database2 – currently estimated to be around 13% of all samples – it is quite possible that two related (or unrelated) individuals share a full SGM+ profile, but the adventitious match has not been identified as it has incorrectly been attributed to duplication.
The official position is: despite the prediction of full false positive SGM+ matches, no such false positive match has been identified.
Current CPS policy dictates that a prosecution will not be commenced on a DNA match alone. This policy affords some protection when an adventitious match has occurred, but may not ultimately prevent an innocent person being prosecuted. In April 1999, Raymond Easton was linked to a burglary that occurred 200km from his house; a crime he could not have committed. The link to the burglary was from an adventitious match obtained with the superseded SGM DNA analysis system. Mr Easton could not have committed the burglary for two compelling reasons: (1) several family members could confirm he was at home caring for his sick daughter the night the burglary occurred; and (2) he suffered from advanced Parkinson’s and could not drive, dress himself or walk more than 10 metres unaided. Nevertheless Mr Easton was arrested, purely on the basis of the DNA evidence, and held in police custody for several hours. It then took a further three months for the charges against him to be dropped. Clearly, Mr Easton’s experience indicates that without the compelling evidence of his innocence, matters may well have proceeded to trial.
I am concerned that SGM+ adventitious matches are in fact occurring but are not being recognised. In recent years there has been considerable focus on obtaining DNA samples from crime scenes in “volume crime” matters (burglary and motor vehicle crime), and so it ought to follow that, if false positive matches have occurred, they are more likely to have occurred in volume crime matters. Due to the CPS policy that a prosecution will not be commenced on a DNA match alone, it is to be expected that an investigation for a minor crime will cease if, at the very least, a link cannot be made between the suspect and the area where the crime was committed. If there is no link and as a result the matter proceeds no further, it is unlikely an adventitious match will be attributed as the reason. If a reason is recorded at all, it is more likely to be put down to ‘insufficient evidence to charge’. However, in the unfortunate case(s) where the adventitious match is to a defendant who lives within a suspicious distance of the crime, a prosecution may well be commenced. Mr Easton serves as an example. Given a defendant is on the database for a reason, bad character evidence may also potentially be available as evidence supporting the prosecution case. When faced with such a scenario a defendant may choose to plead guilty for practical reasons, despite denying it is their DNA at the crime scene. When the DNA link is to a “volume crime”, which will not necessarily carry a lengthy sentence on conviction, challenging the DNA evidence may appear impossible. Alternatively, if a challenge is mounted, particularly in a Magistrates’ Court, it may meet with little success.
Related individuals raise additional concerns, as it can be expected that they will reside close to each other. If one member of a family is on the database but another is not, and those family members share matching profiles, then the person on the database may well end up being investigated and charged for offences committed by their relative. When investigated, particularly where the link is to a volume crime that is not particularly serious, there could be a strong incentive to accept guilt rather than identify their relative. It must also be remembered that individuals may be related without their knowledge, for example where their father is unknown, or incorrectly identified.
Given that further adventitious matches have been predicted, but none have been identified, there must be real concern that an innocent person has been wrongly convicted on the basis of an adventitious match; perhaps because they were not prepared to identify a family member and so entered a guilty plea, or perhaps because unlike Mr Easton they did not have compelling evidence of their innocence. If a miscarriage based on an adventitious match has not occurred yet, it appears likely it will occur in the future as the NDNAD grows ever larger. Greater use of the LtDNA technique will further increase the probability of such a miscarriage.
As a result of my concern about the discrepancy between predicted and identified adventitious matches, I am conducting research into the issue. As part of that research I would be interested in hearing from any defence practitioners who have (a) represented, or are representing, a client where there has been a cold hit DNA match and the client has denied the DNA is theirs, no matter what the outcome of the matter was; and (b) defence practitioners who have represented a client at police interview about a DNA match where no prosecution takes place because no other corroborating evidence exists, and the client’s instructions were that the DNA was not theirs.
I can be contacted by email at: Brian.Costello@1itl.com
Brian Costello graduated with an honours degree in genetics from the University of NSW, Australia in 1996. He worked for five years in genetic research before completing a law degree and being admitted in Australia as a solicitor and barrister in 2004. He was called to the Bar of England & Wales in 2007 and is now completing a third six period of pupillage at One Inner Temple Lane.
