by Sue Carney, ConsultantForensic Scientist, Ethos Forensics
A matching DNA profile provides compelling forensic evidence to the extent that for many, DNA profiling is the gold standard. Police forces request it even in simple cases; the criminal justice system expects it and juries are convinced by it. Advances in DNA technology have made it possible to obtain a DNA profile from the smallest amounts of material; biological traces invisible to the naked eye. That we can detect more from less might be viewed as a useful advance, but what is the value of these low level DNA profiles? Can we reliably identify a match and can such matches provide the same level of evidential significance?
We are all increasingly familiar with the term ‘forensic DNA profiling’. True crime aficionados, armchair detectives and CSI devotees will tell you it’s powerful stuff: DNA profiling can convict the guilty and exonerate the innocent; it has revolutionised forensic science to the extent that an offender’s presence at a crime scene means that his DNA can give him away. Forensic DNA practitioners tell a different story. DNA evidence is not nearly as clear cut as the likes of CSI would have us believe. In reality there are more subtle issues to consider: What is the amount of DNA detected? Can we be sure whether what we’re seeing is DNA from one or more than one person? And, can we be sure of how and when the DNA was deposited? The smaller the amounts of DNA, the more difficult it becomes to reliably answer these questions. And, in those circumstances, the question of whether such DNA results can reliably identify an individual becomes far more subjective than many forensic experts feel comfortable with. Let’s consider the issues in some detail.
Basic Principles and Expectations
Most forensic evidence is based on the principles of transfer and persistence. If traces of evidence transfer, will they persist for sufficient time to be recovered by
investigators? If trace evidence can be recovered, can it be identified? In some scenarios, the answer to each of these questions is a resounding “yes”, and the
context of the evidence, within the framework of case information, allows a relatively straightforward interpretation.
Consider an assault case involving an alleged kicking. A suspect is identified who denies carrying out the kicking assault or indeed, having contact of any sort with the injured man. Contact blood staining is found on the toe of the suspect’s boot and impact blood spatter on the upper. DNA testing of a representative
sample of the blood staining yields a clear, single source DNA profile matching the injured man, who sustained facial wounds that bled heavily during the assault. He states that his assailant knocked him to the ground then kicked him repeatedly to the head and face. In this scenario the blood evidence, in conjunction with the DNA results, provide compelling evidence to support the view that the suspect kicked the injured man, as opposed to the suspect’s version. We can have confidence that the DNA profile is from the blood, and the distribution and location of the blood on the suspect’s boot is an obvious consequence of the alleged action, but cannot be explained by the suspect’s account.
But, what if the trace evidence is not so definitive? Consider a handgun wrapped in a towel found hidden at the back of a cupboard. There is no visible biological material on either the gun or the towel but biological material may be present that is not visible. The towel can be screened using various chemical tests to indicate the presence of other body fluids, but the gun not so readily. Other than a thorough visual examination, the most effective strategy, excluding a separate fingerprint examination, is to swab a selection of the gun’s surfaces and consider DNA testing these samples. What can we expect to detect?
Some contextual information would be useful, for example: Has the gun allegedly been handled by one or multiple individuals? Was the muzzle pushed into someone’s mouth? Did the user strike a person with the gun? The chances are, we don’t have such detailed information, which means our testing strategy is purely speculative. Cost is also an issue. There may not be sufficient forensic budget to DNA test multiple samples from a firearm without further intelligence. In such cases, testing the handgrip and the combined trigger and trigger guard is a common approach.
In our handgun case, It is unlikely, though not impossible, that we will detect a clearly defined, single source DNA profile such as that resulting from the blood tested from the boot. It is far more likely that the resulting DNA profiles will be low level, possibly incomplete (i.e. with components missing) and / or showing indications of the presence of DNA from more than one person, sometimes many people. Our ability to interpret this kind of DNA profile, in particular, to determine whether a reference DNA profile can be said to “match”, is hampered by a number of issues: Low level profiles and mixtures of DNA cannot be attributed to a specific type of biological material; This is likely to cause difficulties in addressing how and when the DNA was deposited; The more complex a DNA mixture, the greater the chance that any randomly selected person’s DNA profile will match by chance, even if they have not contributed DNA to that mixture; The less DNA is present in a sample, the greater the chance that each DNA test from that sample will yield a different profile. This final point is particularly worrying. To be clear: Yes, I am suggesting that there are some circumstances in which two separate tests of the same biological sample can result in two DNA profiles that are significantly different to each other! In order to understand what this means and why it happens, let’s examine the issues in more detail.
Features of a DNA Profile —
How Many Contributors?
Almost all biological organisms on planet earth contain two full copies of DNA. Think of it as an instruction manual written in duplicate: One copy of the manual is inherited from each parent. Each copy is largely the same in terms of overall layout. However, there are some chapters with areas of variation and it’s these differences that can be used to identify individuals in a forensic context.
The UK’s current forensic DNA profiling system examines 16 separate regions of DNA, referred to as loci (singular: locus) plus a sex test. These regions contain the variations mentioned earlier. A DNA profile can be visualised as peaks on a graph; each peak representing a different DNA component, which we call an allele. Each allele is identified by its unique position on the graph and the height of each peak gives an indication of how much of that particular allele is contained in the DNA sample. It is a person’s combination of individual alleles that constitutes their DNA profile.
Which particular alleles are inherited at each locus is completely random. Remember that people have two copies of their DNA, so a parent has two alleles at every locus and it can be either of those alleles that we inherit at that locus. Furthermore, the inheritance of particular alleles at any locus is completely independent of which alleles are inherited at the other loci. This means there are any number of different combinations of alleles that can make up the DNA profile of any person based on their parents’ DNA. Indeed, all of our DNA is inherited in this way, as a random combination of each parent’s DNA, and this accounts for the fact that the population of the planet comprises almost 7.5 billion different people, excluding of course, identical twins (and even they have some small differences).
If by chance a person inherits the same allele from each parent at a particular locus, then we would expect to visualise that as one large peak on our graph, representing a double dose of that allele, described as a homozygous locus. Those loci containing two different alleles are described as heterozygous. Therefore, in a DNA profile representing DNA from one person, there will usually be no more than one or two alleles at each locus. If we see more than two alleles at a locus, the chances are that the tested sample contains DNA from more than one person. There are a few rare circumstances where we might expect to see an extra allele in an individual’s profile but generally speaking, if there are multiple alleles at many loci, we can be sure that there is DNA from multiple contributors.
Determining the number of contributors is important to be able to accurately interpret the DNA results. Depending on the number of contributors, a mixture of DNA can sometimes be resolved into the individual profiles of each contributor, but determining the number of contributors is not as straightforward as you might think. Assessing the maximum number of alleles at a locus can provide some information. In a mixture of DNA from three people for example, we might expect there to be a maximum of six alleles at each locus. However, there are unlikely to be six alleles at every locus. Why? Because one or more of the contributors might have homozygous loci (remember, that’s a double dose of an allele, represented by one larger peak), but also because two or more of the contributors might, by chance, have the same allele in their profile at a particular locus, which would also be represented as a single, larger peak. It’s difficult to determine the proportions of a shared allelic peak that can be attributed to each contributor, especially when we don’t know the individual profiles that are combined to make the mixture, or indeed the amount of DNA that each person has contributed.
There are other problems in trying to determine the number of contributors. What if some of the alleles are not visible? In other words, one or more of our
contributors have provided so little DNA to the mixture that we can only see a part of their profile. How would we know there might be information missing from a profile?
Profile Quality Determines Degree of Interpretation
The possibility of there being missing alleles (or other potential issues) can be assessed by considering the overall quality of a DNA profile. An assessment of
profile quality is linked to the height of the peaks (remember that peak height represents the amount of each allele present in the sample). The smaller the allelic peak heights, the smaller the amount of DNA in the sample. Peak heights above a certain threshold value tell us that there is a significant amount of DNA present, such that we can be confident that there are no missing alleles. In such “good quality” profiles, we can also expect allele pairs (i.e. those at each locus
that are from the same person) to be balanced. In other words, we expect a good quality DNA profile from one person to display the two alleles at each
heterozygous locus at similar heights, reflecting the fact that there are equal amounts of each allele at every locus in that person’s DNA sample.
We can be assured by the properties of good quality profiles in confidently identifying a match when a single source questioned profile contains the same alleles as a reference DNA profile. Further, these properties can be utilised when attempting to resolve a mixture of DNA from two individuals into the individual profiles of its contributors. For reliable mixture resolution, we must first be confident in the assumption that there are two contributors. This is straightforward if we are satisfied that there are no missing alleles. The possible combinations of allele pairs at each locus can then be assessed based on two criteria: 1) the assumption that there is no heterozygous imbalance above an accepted threshold, and 2) the ratio of each contribution of DNA to the mixture, which can be assessed by considering each of the loci that contain the maximum of four alleles. Both of these criteria rely on the fact that in good quality DNA profiles, the peak height data provide an accurate representation of the amounts of each allele within the sample.
We know from studying “poor quality” DNA profiles, i.e. those containing low levels of DNA, that their peak height data cannot provide an accurate representation
of the amount of each allele in the sample. A separation process is required at the later stages of the DNA profiling process in order to determine which alleles are present. Up until that point, all the alleleshad been suspended together in a small plastic tube, each allele having previously been tagged with a luorescent label — copies of the same allele carrying the same tag. When the alleles are separated, a detector measures the fluorescence, which ultimately leads to the identification of the allele. The more copies of a particular allele there are, the more intense the fluorescent signal, the higher the resultant peak representing that allele. The trouble is, in a low level sample, there are fewer copies of each allele, resulting in a low intensity fluorescent signal, which the detector cannot accurately measure. Therefore the height of the resultant peak may not be an accurate representation of the amount of that allele within the sample.
Studies of low level DNA samples and their profiles also show that they are subject to so called stochastic effects that include the possibility of there being missing alleles and considerable imbalance of heterozygous alleles. The DNA profiling process involves a chemical reaction designed to make multiple copies of alleles. The reaction can only produce copies of an allele present in the original DNA extract, which is described as the template DNA. If an allele is not present in the template DNA, then it is not possible for the chemical reaction to reproduce it. Each time the chemical reaction is carried out, a small portion of template DNA is used. The lower the level of DNA in the template, the less likely it is that the portion of template added to the chemical reaction will contain a truly representative selection of all the alleles in the template. In each randomly selected portion of template, some alleles may have been selected in a smaller quantity than is representative of their actual proportion of the template, whilst other alleles may not have been selected at all. Those alleles that were not included in the chemical reaction will not appear in the final DNA profile. This effect is referred to as drop out — the missing alleles having effectively dropped out of the profile. Those alleles selected in a smaller than proportional amount will likely be represented in the final profile as smaller peaks, possibly causing the
profile to display marked imbalance at heterozygous loci. This is compounded by the fact that their peak height data may not be accurate anyway, due to the previously discussed low intensity fluorescence measurement issue. In this situation, the number of possible contributors and the ratio of DNA from multiple contributors cannot be determined from peak heights and heterozygous imbalance rules. It may not be possible to reliably identify a DNA mixture. And, if a mixture is obvious, then it will not be possible to resolve it into its component DNA profiles, based on mixture ratio and imbalance, with any degree of reliability.
There is an additional issue that we haven’t yet mentioned. The increased sensitivity of current DNA profiling techniques means that, in addition to detecting increased numbers of low level and mixed profiles, the technique is also subject to increased risk of contamination. There are many safeguards in place to minimise contamination and to detect it if it should occur, but low level profiles are affected by a particular type of low level contamination referred to as drop in. Despite the best anti-contamination procedures, drop in cannot be completely eliminated from current DNA profiling systems. Imagine a random DNA fragment floating on a minute speck of dust drops into our DNA profiling reaction tube. The chances of this random fragment being an allele are slim. If a person’s entire complement of DNA is described as a large library, then the loci within forensic DNA profiling tests amount to approximately 17 books. In other words, there is far more DNA that is not accounted for by the alleles we are interested in, than DNA that makes up those alleles. If our random fragment is not an allele, then when it drops into our DNA reaction, it will have no affect on the resultant profile whatsoever. However, in the rare circumstance that a drop in is an allele, the reaction will produce more copies of it along with all the other alleles in the reaction and it will appear in the resultant profile and as allelic peak. For all intents and purposes, the drop in is a real allele, it just hasn’t originated from our sample.
The issues of drop in and drop out can be dealt with to some extent by carrying out duplicate testing. Two or more separate portions of DNA extract are added
to two or more chemical reactions, producing two or more separate DNA profiles from the same sample.When the repeated DNA profiles look very different to each other, most likely due to there being a great deal of drop out and imbalance, then this, along with our expectations based on the case circumstances and
sample type, is indicative of a low level result and interpretationis carried out with caution. If an allele appears in one profile but not in the duplicate, then there is the possibility of that allele appearing due to drop in. Because of the rarity of drop in, it’s virtually impossible for exactly the same drop in event to happen twice, so if an allele is appears in both profiles, we can be confident that it is a real allele originating from the sample. In principle, if multiple replicates of DNA profiling tests were carried out, it might be possible to identify all drop in events and to piece together the partial profiles resulting from drop out into one composite
profile. The more replicate profiles, the more reliable the information about the composite profile. This is possible to some degree, but there is usually so little DNA in a low level extract, that there is insufficient template for more than three or possibly four replicate tests.
Increased Complexity Leads to Increased Uncertainty
The increased incidence of DNA mixtures due to the increased sensitivity of modern DNA profiling is, in itself an interpretational issue. We have already discussed
the way in which DNA mixtures from two contributors, and considered to be good quality, might be resolved, based on mixture ratio and the principle of heterozygous imbalance. The solution for two person mixtures is a statistical one, which considers the most likely combinations of alleles at each locus. This statistical calculation is usually carried out by two separate experts who must agree on the allele pairings under consideration, and the evidential significance is then calculated using statistical software. What happens if a mixture of DNA has more than two contributors? Increasing complexity, i.e. increasing numbers of contributors, means that there are more potential allele pairings to consider in resolving the mixture — too many for the analyst to consider manually. There is also likely to be increased sharing of alleles in a complex mixture. This can add an extra layer of complexity to the interpretation, especially if the individuals under consideration are related to each other and might be expected to have more similarities between their DNA profiles than unrelated individuals. Some software packages can separate three and sometimes four person mixtures but these are not yet considered to be routine, although they have been used in some criminal cases. The main issue with increased complexity is the fact that any randomly selected individual’s profile might match some of the corresponding alleles in the mixture by chance, simply because there a so many alleles in the profile. A random “match” in this sense cannot be taken to mean that person must have contributed DNA to the mixture. There is, however, considered to be more significance to such a match when an individual’s reference profile is fully
represented in a mixture, i.e. all of the alleles in the reference profile match corresponding alleles in the mixture.
When considering complex mixtures then, especially if they are also low level, the best we may be able to say is that an individual “could have” contributed their
DNA. The usefulness of “could have” in the absence of a specific strength of evidence is a matter of debate amongst experts. There are widely differing views on
whether it is possible to provide a subjective evaluation of the significance of such a match, in the absence of a statistical value. When do experts ever agree on
anything anyway? In these circumstances though, it’s fair to say that there are multiple factors to consider, and each DNA profile should be assessed on a case by case basis. Legal practitioners will likely be familiar with the ruling in R v Dlugosz, R v Pickering and R v MDS ( EWCA Crim 2), which contains a comprehensive overview of the issues and considers whether such an opinion as “could be a contributor” can be reasonably considered evaluative.
The Bottom Line
In summary then, let’s reconsider the handgun case scenario. The handgrip and trigger / trigger guard samples each produce similar, but not identical, complex
DNA mixtures from multiple contributors — at least four, maybe more. The profile of a suspect who denies any knowledge of the handgun, but who has been inside the property, is compared to each of the DNA mixtures. The alleles in suspect’s profile are found to be fully represented in the DNA mixture from the hand grip and partly represented in that from the trigger / trigger guard. The quality of the mixtures, based on peak height considerations, is not particularly good. There may be drop out since variations between the replicate profiles from each area of the handgun are apparent, and we cannot be sure whether there might also be drop in. Some of the alleles matching the suspect could be described as being the more prominent alleles. However, the mixture is made up from the DNA of a large number of contributors, so we might expect some of the alleles to be shared, which may have contributed to their increased peak height. If our conclusion is that the suspect could have contributed DNA to each of the mixtures obtained from the handgun, then we need to consider how this might impact on the overriding question in this case: Has the suspect handled the gun?
Remember that low level and complex DNA mixtures are under consideration. Even if the suspect has contributed DNA (and we can’t be sure about that, nor can we evaluate our opinion with any strength of evidence), then the DNA results offer no information to suggest what type of biological material the DNA has originated from, how, or indeed, when it was deposited.
The expected result from a handled item, compared to a sample of visible body fluid, is a low level DNA profile. Indeed, many practitioners refer to this type of result as “Touch DNA”, which may be erroneous. Consideration must also be given to other, indirect mechanisms of DNA deposition. The towel, found wrapped around the handgun, has not yet been DNA tested (and even if it were to be tested, it is a large item, with no features to suggest which areas should be sampled). The suspect does not deny being at the property. It’s possible that on some previous occasion the suspect used the towel and it was later wrapped
around the handgun, at which point some of his DNA was transferred to the firearm. There are extensive data in the scientific literature that demonstrate the
secondary transfer of DNA from one surface to another, via an intermediate surface. It’s also possible for the towel to have rubbed off the DNA of the most recent user of the handgun. There are a multitude of possibilities based on other pieces of information that have not yet been suggested and whilst it is not feasible to consider them all, the DNA expert must acknowledge the uncertainty surrounding how and when the DNA was deposited.
Returning to the original premise then — that advances in the sensitivity of DNA profiling are advantageous; that a DNA “match” provides compelling evidence — it’s clear that this is not always the case. DNA evidence may be complex and must be considered in the light of the case contextual information. Particularly if DNA profiles are low level, it may not be possible to evaluate the significance of a match and it is unlikely that such a match will assist in answering questions about how and when the DNA was deposited. Assumptions about a DNA match in this context may be unsafe. The significance of each match must be considered with caution and its limitations properly explained to the court if we are to avoid conclusions based on misconceptions and to prevent miscarriages of justice.