Can we rely on the DNA match statistic from specialist statistical software?
As DNA profiling methods have become ever more sensitive, the detection of trace amounts of DNA has significantly improved. As a result, the number of DNA mixtures we encounter has increased. This was especially highlighted following the introduction of DNA17 methods in 2014. In many situations the mixtures have multiple contributors and are complex to interpret. Whilst previously many results were designated as too complex for interpretation, the last few years have seen the introduction of ‘probabilistic genotyping software’ packages which enable a statistical evaluation of complex results.
Some of these software packages are available commercially whilst others are open source or in-house only; and their underlying methodology and approach varies. For example, Key Forensic Services uses STRmix® a commercially available package developed in New Zealand. Cellmark Forensic Services uses both STRMix® and an in-house solution ’DNA Resolve’ based on an open source software package (LikeLTD) developed by Professor David Balding; Eurofins Forensics (formerly LGC Forensics) uses LiRa, a program developed in-house.
What are the issues?
- There is an ongoing debate within the scientific community as to which program is best. Some use more information from the DNA result than others (e.g. presence/absence of DNA components, size of the DNA components) and they treat ‘uncertain’ alleles in different ways. The DNA expert must still make some subjective decisions about the DNA result before the data is processed by the software.
- If the same DNA result was put through each of the software programs, a range of likelihood ratios would be obtained. If the software is robust, ideally they should all be in the same ballpark; it has been suggested that results within one ‘ban’ (a factor of 10) of each other is acceptable. However, there is no ‘Gold Standard’ package and no ‘right’ answer.
- Many of the software packages are validated on data specific to the organisation using the program, so an expert instructed by the defence wishing to independently repeat the calculations requires access to the software program and the data used by the Crown for validation.
- Different programs can deal with different numbers of contributors but some results will still be too complex to interpret.
- Where it is not possible to provide a statistical evaluation, there is no objective measure of how much more likely it is that a particular individual has contributed DNA to that mixture rather than their DNA profile matching by chance. In this situation, in our view the result should be considered as inconclusive.
- The UK Forensic Science Regulator has published guidelines for the interpretation of mixtures and the validation of the software.
- Whilst many DNA experts have an understanding of the programs and can review whether a robust set of data from the DNA result has been input to the software, any in depth challenge to the underlying statistics is likely to require input from a statistician. There are very few reputable experts who do this. However in many cases, especially when low levels of DNA are present, it is the means of transfer of the DNA that is the most crucial issue in the case – there may be a legitimate reason for the transfer, unrelated to the crime.
- It is always worth the defendant instructing an expert to review complex and low level DNA findings, especially when there may be alternative opportunities for DNA transfer.
Forensic Context’s experts can undertake a review of complex DNA match statistics (likelihood ratio) provided by the new software programs. Such reviews will inform the client as to whether the figures quoted are robust, fair and reasonable. Most importantly we will review the DNA match in the context of the case scenario, and taking into account both the match statistic and the transfer issues, evaluate how likely the DNA findings are given the defendant’s account compared to that of the complainant. Many Crown statements provide only the DNA match statistic which has the potential to mislead the court if they do not realise that this provides no information as to how or when the DNA could have been transferred.
In a murder case review carried out by one of our experts, the Crown obtained a DNA result from the neck swabs of a man who had allegedly been strangled by the defendant. The result was weak, low level and a mixture. The major (stronger) part of the result matched the deceased, which was not unexpected given that the swabs had been taken from his neck. However the minor (weaker) part of the result was of interest and had been interpreted by the Crown scientist with a view to assessing whether or not the defendant could have contributed.
There were indications of at least two contributors to the minor part and the Crown scientist’s statistical evaluation was carried out on this basis. Their view was that the result was 130,000 times more likely if the defendant had contributed rather than if he had not. However if all of the uncertain DNA components in the result were taken into consideration, there were indications of DNA from up to 5 individuals in the minor part of the result, a scenario which was far more complex and of which the Crown’s calculation did not appear to adequately take account. Furthermore, possible indirect transfer mechanisms for the presence of the defendant’s DNA on the neck swabs were also evaluated by our expert. The Crown’s scientist agreed with our expert that the DNA findings did not favour primary transfer over indirect transfer. The defendant was acquitted.