The State of Western Australia v Mitchell [No 2]
[2018] WASC 29
•29 JANUARY 2018
JURISDICTION : SUPREME COURT OF WESTERN AUSTRALIA
IN CRIMINAL
CITATION: THE STATE OF WESTERN AUSTRALIA -v- MITCHELL [No 2] [2018] WASC 29
CORAM: HALL J
HEARD: 15-18 MAY & 12 JUNE 2017
DELIVERED: 17 AUGUST 2017
PUBLISHED: 29 JANUARY 2018
FILE NO/S: INS 152 of 2015
BETWEEN: THE STATE OF WESTERN AUSTRALIA
Prosecutor
AND
ERNEST JAMES MITCHELL
Accused
Catchwords:
Criminal law - Evidence - Admissibility of expert opinion evidence - DNA evidence - Whether DNA evidence unreliable due to the possibility of contamination - Whether errors made in assessing number of contributors to a DNA profile resulting in inaccurate assessment of probability that accused is a contributor - Whether methods used by laboratory had been validated - Whether proficiency of scientists established - Turns on own facts
Criminal law - Evidence - Whether discretion to exclude otherwise admissible evidence on the grounds of unfairness should be exercised - Turns on own facts
Legislation:
Nil
Result:
The evidence is admissible
Category: B
Representation:
Counsel:
| Prosecutor | : | Mr J C Whalley |
| Accused | : | Mr J A Davies |
Solicitors:
| Prosecutor | : | Director of Public Prosecutions (WA) |
| Accused | : | Gary Rodgers Barrister & Solicitor |
Case(s) referred to in decision(s):
Aytugrul v The Queen [2012] HCA 15; (2012) 247 CLR 170
Dasreef Pty Ltd v Hawchar [2011] HCA 21; (2011) 243 CLR 588
Her Majesty's Advocate v Sinclair [2014] HCJAC 131
Liyanage v The State of Western Australia [2017] WASCA 112
Pantoja (1996) 88 A Crim R 554
R v Dlugosz [2013] EWCA Crim 2
R v Dos Santos (Unreported, London Central Criminal Court, March 2012)
R v Pantoja (1996) 88 A Crim R 554
R v Reed and Reed, R v Garmson [2009] EWCA Crim 2698
R v Reid [2009] NZCA 281
R v Wallace [2010] NZCA 46
State of WA v Wark [No 2] [2018] WASC 18
Tuite v The Queen (2015) 49 VR 196; [2015] VSCA 148
HALL J:
Summary
The accused is charged that on a date unknown between 8 January 2009 and 15 January 2009 he murdered Robert George Dalliston. He has entered a plea of not guilty to that charge and the matter is set down for a trial commencing on 7 March 2018.
By an application dated 23 April 2017, the accused sought an order that prosecution expert evidence of DNA analysis be ruled inadmissible. The DNA evidence in issue relates to the testing of samples taken from a number of items at the scene of the alleged murder. Four of the samples produced complex mixed DNA profiles, meaning that in each case the result was consistent with multiple contributors and that the DNA was present at comparatively low levels.
The samples have been the subject of analysis at laboratories in this State, South Australia and the United Kingdom. That analysis has included (in the case of South Australia and the United Kingdom) the use of statistical software programs specifically developed for the interpretation of low level mixed samples. The results of that analysis are likelihood ratios that in each case favour a hypothesis that the accused is a contributor to the mixed samples of DNA, to varying levels of probability. This evidence forms a significant component of the State's circumstantial case against the accused.
The State will rely on the DNA results as evidence that the accused was present in the house in which the deceased was murdered. This is contrary to his assertions to the police when interviewed that he had never been inside the house. The State is likely to rely on this denial as a lie arising from a consciousness of guilt. This represents one of the principal strands of the prosecution case. The other principal strand is an alleged admission made to a third person.
The application to exclude the evidence is based on three grounds:
1.The results are said to be unreliable due to the possibility of contamination.
2.Errors have been made in assessing the number of contributors to the mixed samples and that this has resulted in inaccurate assessments of the probability that the accused is a contributor.
3.There has been no, or insufficient, proficiency testing or validation studies to establish the reliability of the probability calculation methods for low‑level, complex mixed DNA samples used in this case.
The evidence in issue is that of four forensic scientists. They are Mr Scott Egan of PathWest, Dr Duncan Taylor of Forensic Science South Australia (FSSA), and Mr Huw Turk and Dr Michael Walbank, both of Cellmark Forensic Services in the United Kingdom (Cellmark). Each has produced witness statements, reports and copies of documents relating to the analysis they have undertaken.
A directions hearing to determine the admissibility of the evidence was conducted between 15 and 18 May 2017. At that hearing, the four prosecution witnesses were called and cross‑examined at length. In addition, the defence called their own expert witness, Dr Brian McDonald. Written submissions were filed by both the prosecution and the defence. Final oral submissions were made on 12 June 2017.
On 17 August 2017, I ruled that the evidence is admissible and also declined to exercise the discretion to exclude otherwise admissible evidence on the grounds of unfairness. I said that detailed reasons for my decision would be provided to the parties. These are those reasons.
In these reasons I will deal with the following matters:
1.The general nature of the evidence in issue.
2.Relevant principles of law.
3.The defence contentions.
4.The prosecution contentions.
5.The evidence given in regard to the issues of:
(a)contamination;
(b)contributor numbers; and
(c)proficiency and validation.
6.Conclusions.
The evidence in issue
In order to understand the significance of the evidence in dispute, it is necessary to first briefly summarise the prosecution case.
At the time of his death, the deceased, Mr Dalliston, was a 69‑year‑old invalid pensioner. He used a walking stick and lived alone at a house in Mandurah. The accused lived in the same street. Sometime between 8 January 2009 and 15 January 2009 the accused is alleged to have gone to the rear of Mr Dalliston's house, gained entry through a window and killed Mr Dalliston using a concrete slab and the walking stick.
The body of the deceased was not found until 15 January 2009, by which time it was in a badly decomposed state. A large number of swabs were taken from the scene for the purpose of DNA analysis. Four of the swabs ultimately produced results that are relied on by the prosecution and that are the subject of this application. They are swabs taken from a wallet, a doorknob, a thong and a bedsheet. The samples in question will be referred to by reference to those items throughout these reasons.
A fifth result was obtained from a sample taken from a window frame, but that result was subsequently withdrawn and is not relied upon by the prosecution. It will, however, be necessary to refer to the window frame analysis in assessing the adequacy of procedures at PathWest.
Initial testing of the exhibits in this case, of which there were over 1,500, was carried out by PathWest. Subsequently, the bedsheet result was sent to FSSA for further statistical analysis and the other three samples were sent to Cellmark for both further testing and analysis. As with all such testing, a kit is used that identifies the components (alleles) present at a number of sites (loci) on the DNA molecule. Chosen sites are ones known to have a high degree of variability between individual people. A number of different kits were utilised, which vary as to the number and location of the sites examined. At each site, an individual will normally have two components, one inherited from each parent.
The initial testing by PathWest used the Profiler Plus DNA profiling kit. That kit tests 10 sites. In 2013, the more sensitive PowerPlex 21 profiling kit was introduced at PathWest. That kit tests 21 sites. Many exhibits were then retested by PathWest. None of the testing undertaken by PathWest produced a result that was capable of further analysis by that laboratory. However, there were limitations on the ability of PathWest to conduct statistical analysis of some of the samples. It was for this reason that FSSA and Cellmark were engaged.
The PathWest testing remains relevant because contamination was detected in some of the negative controls used in that testing. Each sample being tested is accompanied by a negative control as it passes through the various stages of testing. The purpose of the negative control is to identify any contamination risks within the laboratory. Contamination in a negative control raises the possibility that the associated test sample may also be contaminated. The contamination identified by PathWest was investigated and it was determined that it did not affect the validity of any reported results (other than in the case of the window frame).
The PathWest testing of the bedsheet is also relevant because that testing was relied upon by FSSA when undertaking further statistical analysis. The Profiler Plus result in respect of the bedsheet was assessed as being a mixed sample from at least two contributors and indicated that the deceased and the accused were possible contributors. At the time, PathWest was not accredited to use the relevant software (STRmix) to produce a likelihood ratio for a mixed sample derived from Profiler Plus. FSSA did have the relevant accreditation. This is why the results were sent to FSSA for further analysis by Dr Taylor.
The other three samples were sent to Cellmark in 2016, both for further testing and analysis. Cellmark used a different DNA profiling kit, NGM Select. This kit tests 17 sites. Cellmark also tested the related negative controls used at PathWest and detected some contamination. That contamination was investigated and it was determined that it did not affect the validity of the reported results. Cellmark then used Like LTD, a probabilistic software program similar to STRmix, to produce likelihood ratios in respect of the wallet, doorknob and thong samples. The essential difference between STRmix and Like LTD is that the former takes into account peak heights, whereas the latter does not.
In order to conduct a statistical analysis, it is necessary for a scientist to make an assessment of the minimum number of contributors that could explain the DNA profile obtained. This requires examination of an electropherogram, a chart that shows peaks representing the alleles present at each of the test sites. Because each individual will normally have two alleles at each site, the presence of more than two peaks will indicate the presence of more than one contributor. The number of contributors is not, however, simply a matter of dividing the number of peaks by two. An individual can inherit the same allele at a particular site from both parents and this will produce a single peak at that site, though it may be higher because that allele is present in a greater quantity. Further, in a mixed sample it is possible that two or more individuals can have a common allele at a particular site. It is also possible that peaks may appear that are artefacts of the testing process and do not represent the presence of alleles in the sample. In making an assessment of the number of contributors, the scientist needs to exercise judgment as to which of the peaks should be counted and which discounted.
All of the relevant samples were mixed; that is, the results in each case were consistent with more than one person contributing to the DNA that was found. In each case, a confirmed sample of DNA from the accused was compared to the mixed samples and an analysis done to determine the likelihood of the accused being a contributor to the mixed sample. As is usual in such cases, the results are expressed as a likelihood ratio. This ratio compares two competing hypotheses. The first hypothesis is that the accused and unknown unrelated others are the contributors to the mixed sample. The second hypothesis is that the accused is not a contributor and that the DNA comes from other unknown and unrelated people.
The results in issue are as follows:
1.Bedsheet - the mixed DNA profile is 53 million times more likely to have been obtained given the hypothesis that the accused is a contributor than the hypothesis that he is not a contributor.
2.Wallet ‑ the mixed DNA profile is approximately 72,000 times more likely if the accused is a contributor than the hypothesis that his DNA profile components are represented by chance.
3.Doorknob - the mixed DNA profile is approximately 480,000 times more likely if the accused is a contributor than the hypothesis that his DNA profile components are represented by chance.
4.Thong ‑ the mixed DNA profile is approximately 13,000 times more likely if the accused is a contributor than the hypothesis that his DNA profile components are represented by chance.
The reason for the slightly different wording is that FSSA and Cellmark express their conclusions differently. There is no significance in this difference.
General principles
Expert opinion evidence on a scientific matter will be admissible if an affirmative answer can be given to the following four questions:[1]
1.Is the opinion relevant, ie, could the evidence rationally affect, directly or indirectly, the assessment of the probability of the existence of a fact in issue in the proceedings?
2.Is a person of ordinary experience unable to form a sound judgment on the subject matter without the assistance of an expert witness with special knowledge or experience in the area?
3.Is the subject matter part of a body of knowledge or experience that is sufficiently organised or recognised to be accepted as a reliable body of knowledge or experience?
4.Has the witness acquired, by study or experience, sufficient knowledge of the subject to render his or her opinion of value in resolving the issues before the court?
[1] See Liyanage v The State of Western Australia [2017] WASCA 112 [122].
In this case there is no issue that the evidence in issue is relevant, outside the experience of ordinary people and that each of the prosecution witnesses is qualified to give expert evidence. It is also accepted that DNA analysis in general is an accepted and reliable body of knowledge. The issues are whether contamination has rendered the particular results in this case unreliable, whether critical errors (or unjustified assumptions) were made in the assessment of contributor numbers, and whether the methodology for testing and analysis of low-level mixed samples is sufficiently reliable. The last of those issues raised the use of probabilistic computer programs, of which both STRmix and Like LTD are examples. The validity of STRmix was accepted, but that of Like LTD was challenged. There was also a question of whether laboratory processes for the analysis of low level DNA samples had been validated and the proficiency of the scientists established.
The validity and reliability of the STRmix program was considered by the Victorian Court of Appeal in Tuite v The Queen.[2] Though STRmix is not challenged here, what was said in that case is also relevant to Like LTD. In that case there was a challenge to the use of STRmix because it was said to be a discrete or new area of knowledge that was yet to achieve general acceptance in the scientific community. The court agreed with the trial judge that STRmix and the methodology it uses is a development within an established and sophisticated field of knowledge concerned with the evaluation of DNA profiles. Rather than being a new or discrete body of knowledge, it is a new development within an existing body of knowledge.[3] The court quoted the trial judge's reasons with approval, in particular:
In my view, this evidence places STRmix and its fully-continuous probabilistic methodology squarely within the field of statistical DNA evaluation, albeit possibly at its leading edge. It has a credible scientific and mathematical basis. However, that is not to say that the results that it produces are inherently reliable. It may not produce results that are as reliable for the purposes of forensic case-work as binary or semi-continuous methods. Its method for determining the possibility or probability of drop-out based on peak height variability may be open to criticism. Its use on compromised profiles may be questionable. But those are matters for competing expert opinion, providing, of course, that STRmix is amenable to scrutiny and independent testing…
I have found the STRmix methodology to be a development in a recognised field of knowledge concerned with the statistical evaluation of DNA profiles. It is not subjective or speculative or otherwise to be dismissed as lacking an objective basis. In my view, based on the opinions of Dr Taylor, Ms Federle and Ms Scott the limitations identified do not substantially erode the probative worth of the DNA evidence.
However, it is clear that there is scope for competing expert evidence about the reliability of the STRmix methodology. What a lack of international take-up or independent review and assessment of the STRmix methodology means for the weight that should be given to the DNA evidence will be a matter for the jury, based on differing expert evidence on the issue. Likewise, the suitability of STRmix (or any system based on peak-height modelling) for the analysis of low-template DNA will be the subject of disagreement between experts upon which the jury will be called upon to adjudicate [116] ‑ [122].
[2] Tuite v The Queen (2015) 49 VR 196; [2015] VSCA 148.
[3] Tuite [116].
The views expressed in Tuite were based on the evidence adduced in that case (including from Dr Taylor, who also gave evidence on this application). Any conclusions in this case are similarly dependent on the evidence. It is relevant, however, to note that another court has accepted that probabilistic methods of analysing mixed DNA samples are a development within an existing accepted field of expertise. It is also relevant that competing expert views as to reliability were viewed in that case as being issues for the jury to consider in determining the weight of the evidence.
Similar views about the relevance of reliability were expressed by the New South Wales Court of Criminal Appeal in R vPantoja.[4] That was a case involving DNA evidence where a defence expert had given evidence that on his testing the accused was excluded as a contributor to the sample, evidence that contradicted prosecution witnesses. Hunt CJ at CL said that conflicts between scientific witnesses are to be resolved by the jury, not the trial judge. He noted that any reasonable doubt arising from the conflict in the expert testimony in that case was not such that it was incapable of being safely eliminated.[5] Abadee J expressed a similar view, saying that it was for the jury to resolve conflicts in the expert evidence, including as to whether proper techniques had been used and whether procedures were adequate and reliable. Where such issues are raised, a jury should be directed that they should disregard the evidence if they do not accept that the methodology on which it is based is sufficiently reliable, accurate and dependable.[6]
[4] R vPantoja (1996) 88 A Crim R 554.
[5] Pantoja (559).
[6] Pantoja (576 - 577).
As to the reliability of DNA tests where the sample is small, evidence regarding so-called Low Copy Number (LCN) or Low Template DNA has been considered in a number of cases. It was held to be admissible in R v Reid,[7] and in Wallace v R.[8] The court in Wallace observed that the accuracy of the results generally reduces as the amount of DNA reduces.[9] In normal DNA processes, amplification of the DNA is achieved by copying it using 28 cycles. In the LCN process 34 cycles are used. This technique is used where the amounts of DNA involved are very low.[10] Though the amounts in this case were small, the available specialist high-sensitivity technique was not used.[11] Such techniques may have the effect of amplifying stochastic effects and for this reason a threshold quantity may be suggested. It was not suggested that any threshold is applicable here. Nonetheless, some of the problems associated with interpreting results from very small samples of DNA are also applicable to the samples examined in this case.
[7] R v Reid [2009] NZCA 281.
[8] Wallace v R [2010] NZCA 46.
[9] Wallace [111].
[10] See Wallace [94].
[11] Statement of Mr Turk, exhibit P1, page 41.
These issues were considered by the UK Court of Appeal in R v Reed and Reed, R v Garmson.[12] In that case, the court concluded that LCN DNA can be used to obtain profiles capable of reliable interpretation provided that the profile is at a level where it is unlikely to suffer from stochastic affects which would prevent proper interpretation. The court anticipated that there would be cases where there was disagreement as to the impact of stochastic effects and, in such cases, expert evidence must be given as to whether a reliable interpretation can be made.[13] Later, the court expressed the view that there is no enhanced test for the admissibility of such evidence. The court said:
If the reliability of the scientific basis for the evidence is challenged, the court will consider whether there is a sufficiently reliable scientific basis for that evidence to be admitted, but, if satisfied that there is a sufficiently reliable basis for the evidence to be admitted, then it will leave the opposing views to be tested at the trial [111].
[12] R v Reed and Reed, R v Garmson [2009] EWCA Crim 2698.
[13] R v Reed and Reed, R v Garmson [74].
See also R v Dlugosz .[14]
[14] R v Dlugosz [2013] EWCA Crim 2.
It is also necessary for an expert to identify any facts relied on, assumptions made and the course of reasoning that has been followed.[15] The accuracy of the assumptions and the reliability of the methodology can then be tested. The fact that these matters are able to be challenged, or even disputed by other experts, does not, however, necessarily mean that the evidence is inadmissible. It would only be if the evidence on a voir dire established that the methodology was incapable of acceptance by a jury that the evidence would be inadmissible.
[15] Dasreef Pty Ltd v Hawchar [2011] HCA 21; (2011) 243 CLR 588.
As to the risk of unfairness, the High Court considered this in the context of evidence regarding mitochondrial DNA in Aytugrul v The Queen.[16] In that case, the opinion of the expert had been expressed using a percentage figure and it was argued that there was a risk that the jury would round up a high percentage likelihood to 100% certainty. In their judgment the plurality said that in assessing the danger of unfair prejudice regard must be had to all of the evidence that is to be given, not merely to the results.[17] Their Honours then said:
Evidence given about the results of DNA analysis is evidence about comparisons between identified samples and one or more databases. The results of those comparisons can be expressed qualitatively or quantitatively. If expressed quantitatively there are assumptions and approximations made which often (perhaps always) require elucidation and explanation to make plain what are the limits to the opinion that is being expressed as a number or range of numbers…it will usually be important, even necessary, that the evidence provides the jury with so much of the expert's 'specialised knowledge' as the jury requires properly to understand the opinion expressed ‑ and what it can and cannot demonstrate ‑ and that this specialised knowledge be related to the facts of the case [32].
[16] Aytugrul v The Queen [2012] HCA 15; (2012) 247 CLR 170.
[17] Aytugrul [30].
Whilst the views in Aytugrul were expressed to be in the context of the Uniform Evidence Act, the point that the risk of unfair prejudice must be considered in the context of the evidence as a whole is equally applicable to all cases.
The defence contentions
In the present case, it is accepted on behalf of the accused that DNA as an evidentiary tool has been in common use for many years and that it is a specialised area of knowledge that is often the subject of admissible expert evidence by suitably qualified scientists.[18] The accused did not challenge the qualifications of any of the prosecution witnesses. The relevance of the evidence is also not in dispute.
[18] Defence supplementary submissions, dated 2 May 2017, par 26.
There is, however, an issue as to whether the samples in relation to which the likelihood ratios were made were contaminated. It is submitted that the existence of contamination in some of the negative samples means that there is a distinct possibility that the crime scene samples were also contaminated. The defence asserts that in these circumstances the samples should have been rejected and any results based on them are unreliable. Even if this is not sufficient to make the evidence inadmissible, the defence submits that it justifies excluding the evidence in the exercise of the court's discretion. The argument is that no jury would be capable of taking into account questions as to the reliability of the processes used because they would be over-awed by the statistical results. I will refer to contamination as issue one.
There is also an issue as to the assessment of contributor numbers. The defence position is that the assessment of peak heights and contributor numbers in low level mixed samples is uncertain and subjective, and that no, or no adequate, verification that such results are reliable has been undertaken. The objection was focussed on whether the testing results in this case were suitable for interpretation. That is, the defence contention is that the low levels of DNA in this case were not a proper basis for statistical analysis because there was too much uncertainty as to what the test results showed. It was also suggested that the contributor numbers assigned were capable of dispute and that each sample could have involved higher numbers. I will refer to contributor numbers as issue two.
Finally, the defence suggests that the laboratories involved used methods that had not been validated and that the proficiency of the scientists was not established. In particular, it was submitted that the Like LTD software program that was used to generate the likelihood ratios for the three samples analysed by Cellmark is insufficiently validated. By this I understood the defence to mean that whilst the use of statistical programs to analyse DNA profiles has been in use for some time, this particular program is not one that has been shown to be reliable or that attracts uniform support or endorsement from other scientists. No similar claim was made in respect of STRmix. I will refer to this as issue three.
The Prosecution Contentions
The prosecution submits that the defence contention regarding contamination demonstrates a poor understanding of the laboratory procedures and the potential significance of the contamination. Contamination of a negative control requires investigation. It does not necessarily mean that the sample is also contaminated. In this case proper investigations were undertaken. There was no contamination in respect of the bedsheet and contamination in respect of negative controls for the other items was taken into account. The likelihood ratios were not affected. The State says that Dr McDonald lacked objectivity and made a number of errors in his reports. It would be open to a jury to find that the crime scene samples were not contaminated and that reported likelihood ratios are reliable.
The State accepts that assessment of contributor numbers is necessary for the calculation of a likelihood ratio. Whilst there is no way of ever knowing the number of contributors with absolute certainty, an assessment can be made by a suitably trained and experienced scientist of the minimum number of persons likely to have contributed to the sample. The defence had not raised any issue with the assessment of contributor numbers by Dr Taylor in respect of the bedsheet. As to the other samples, they were examined by Cellmark and familiarity with the particular test kit is of importance. Dr McDonald was not familiar with the NGM Select kit, unlike Dr Walbank and Mr Turk. At its highest, the defence claim that the selection of contributor numbers by Cellmark is unreliable is a difference of scientific opinion as between Dr McDonald and the Cellmark scientists.
As regards validation and proficiency, the State notes that it only intends to lead evidence of results produced by FSSA and Cellmark. However, the State nonetheless maintains that PathWest possessed the necessary accreditation and proficiency to effectively carry out the testing. The sole basis for objection to Dr Taylor's evidence appears to be that the likelihood ratio he produced using the Profiler Plus results is different from one that was generated by PathWest using PowerPlex 21 ‑ but this difference is explained by the use of the different kits. In regard to Cellmark, it is accredited to use the Like LTD software, a system that has been comprehensively validated and reviewed. As to the claim that Cellmark scientists have not had proficiency testing, the State says that this is simply wrong and that the defence have been provided with access to the test results. Dr McDonald's opinion that the relevant standards have not been met is something that should be ventilated before a jury.
The evidence at the directions hearing
For the purposes of the directions hearing, the parties agreed to the tender of a book of materials. That book contained copies of the statements of the witnesses called at the directions hearing, academic articles and other documents referred to by the witnesses or relied upon by the parties. Each of the prosecution witnesses was called to give oral evidence and the defence called Dr McDonald.
It is neither necessary, nor appropriate, to make any assessments of credibility in respect of the witnesses. The present task is to determine the admissibility of the evidence. In the event that it is admissible, it is for the jury to determine what weight should be given to the evidence. I do, however, note that Dr McDonald candidly admitted that he had no recent laboratory experience, no experience in the production of low-level mixed profiles, no experience in the use of NGM Select and no experience in the use of the probabilistic analysis programs (STRmix and Like LTD) utilised in this case.[19]
[19] ts 446.
It is convenient to summarise the evidence given by the witnesses by reference to the issues raised in the application.
Issue 1 - contamination - general
Mr Egan confirmed that in addition to its own protocols and procedures, PathWest is accredited by the national accreditation authority, the National Association of Testing Authorities (NATA). One aspect of this is an assessment every 18 months to two years to ensure compliance with national standards. PathWest has maintained its accreditation for the last 17 years.[20]
[20] ts 200.
Reference was made to an employee who had been dismissed for breaches of procedure, in particular failing to have results peer reviewed and failing to undertake proficiency testing. That person had some early management involvement in the case but no involvement in the actual testing of items.[21] Mr Egan rejected any suggestion that contamination events in this case, when taken together, showed that there had been systemic failure in quality controls at PathWest. He said that no concerns in that regard had been raised by NATA.[22]
[21] ts 201 - 203.
[22] ts 203 - 204.
In cross‑examination, Mr Egan was asked about an error that had occurred in 2013 involving a positive control sample. Positive controls are used to ensure that the amplification processes operate properly, but the profile generated by the positive control should not be loaded onto the database. This had occurred in 2013 and was not detected until 2016. It was suggested that this showed a deficiency in the quality control systems.[23] A second instance had occurred in 2009 when a staff member's DNA had inadvertently contaminated a sample and had been loaded onto the database. Again Mr Egan accepted that this was regrettable, but not unexpected in a laboratory that deals with many thousands of samples each year. Mr Egan accepted that these events had not been immediately detected, but that it was in the nature of quality control systems that they needed to be constantly improved and that measures had been put in place to prevent this occurring again. He said that it was unfair to cite two instances as evidence that the system is a failure, when in fact the quality assurance system functions very well.[24] I note that it was not suggested that either of these events had any direct potential to affect the results in this case.
[23] ts 213 - 214.
[24] ts 215 - 216.
Reference was also made in cross‑examination to a proficiency test in which possible contamination had been detected. Like all accredited laboratories, PathWest undertakes regular proficiency testing. This involves known samples of DNA being sent by an independent testing service, analysed by PathWest and the results then checked. In a test undertaken in 2013, in addition to the intended sample DNA, some extraneous low‑level DNA was detected. It was suggested that this was evidence of contamination at PathWest. Mr Egan said that there were a number of possible explanations. The sample could have been contaminated when it arrived. Furthermore, only five extraneous peaks were observed, four of which could be explained as being artefactual (products of the processing system) leaving only one, that could be explained by drop‑in of a stray allele from the environment. None of this suggested a failure of the system.[25]
[25] ts 219 - 224.
Each of the three events just referred to was also referenced by Dr McDonald in his evidence. He said that they showed a failure on the part of PathWest to give proper attention to contamination risks and to respond to contamination events when they occurred. He said that given that the samples were at low levels, other instances of low level contamination in the laboratory made it too dangerous to rely on the results in this case.[26]
[26] ts 444 - 445.
Mr Egan produced a microtiter plate and aseptor strip (exhibit P4) to demonstrate how samples and controls were placed into the electrophoresis machine. The microtiter plate is a plastic tray in which there are 96 small wells. A sample of control fluid is placed into each individual well to a maximum of one third capacity. This is done using a needle or glass tube, which must pass through the aseptor strip. The aseptor strip is a small rubber mat with plugs that seal each well. The purpose of the aseptor strip is to minimise evaporation and prevent contamination. Each plate and strip is used only once and then discarded. A negative control is matched to each sample as well as there being an additional control introduced at each stage of the process. In this case no contamination was detected in the amplification stage, only in extraction negative controls.[27] In cross‑examination, Mr Egan accepted that this process minimised but did not entirely prevent the risk of contamination.[28] Dr McDonald gave similar evidence.[29]
[27] ts 408 - 411.
[28] ts 412.
[29] ts 435.
Dr Walbank emphasised that each sample tested at Cellmark was subjected to multiple analysis runs. Though only one result is reported, it is a report based on the consensus result of multiple tests, which means that the result is more robust.[30]
[30] ts 501.
Dr McDonald was of the view that the contamination events in this case were of sufficient seriousness that the results in respect of the samples should not have been relied on. He said that together they showed systemic failure of quality controls.[31] However, he accepted in cross‑examination that contamination of a negative control does not necessarily mean that the associated sample is contaminated. He also accepted that there was no evidence of contamination in the bedsheet negative control or the thong negative control and that the peaks in the doorknob control did not appear in the sample.[32] He accepted that PathWest investigated each of the contamination events, though he considered that they were not done in a timely way and he disagreed with the conclusions.[33] He accepted that it was possible that contamination in the negative controls did not affect the likelihood ratios that were reported.[34] He accepted that he had made errors in his report as to the number of contributors that had been assumed for some of the calculations.[35] He said that his report had not been peer reviewed.[36]
Issue 1 - contamination - the bedsheet sample
[31] ts 437, 452,
[32] ts 447 - 449.
[33] ts 450 - 451.
[34] ts 453.
[35] ts 455 - 456.
[36] ts 457.
Mr Egan gave evidence that no contamination was detected in the negative control relating to the bedsheet, either in 2009 using Profiler Plus or in 2013 using PowerPlex 21.[37] However, at each stage of testing, the relevant negative control may be batched with controls for other samples. That is, a number of controls may be included on the same tray or be in adjacent containers. Low level DNA was detected in 2013 in another negative control (4422) in the same batch run as the bedsheet control. That contaminated control related to the window frame and resulted in the report for the window frame being later withdrawn.[38]
[37] ts 196 - 197.
[38] ts 197 - 198.
Mr Egan said that a negative control has two main functions - to test the validity of the chemicals used and to identify if there is any cross‑contamination. Contamination may vary in extent from a single allele to a full profile. Contamination in the negative control does not necessarily mean that the associated sample is contaminated. If contamination is detected, an investigation is conducted and the run is then repeated to see if the same result is produced. If the contamination result is repeated, the result is not cleared for reporting. PathWest has a policy and guidelines for dealing with contamination events.[39]
[39] ts 190 - 191.
Mr Egan was asked in cross‑examination whether an unrelated contaminated control in the same batch should invalidate the whole batch. He said that was not so because a related control follows a sample through the process and if there are no problems with that control at any stage then that should validate the sample.[40] Dr McDonald was of a different view.[41]
[40] ts 227 ‑ 228 and 279.
[41] ts 342.
As noted earlier, another sample, from a window frame, produced a positive result, but this result was later withdrawn by PathWest. The reasons for that are relevant because the prosecution submits that they show the conservative approach taken by PathWest to issues of contamination. When the window frame was re-tested by PathWest in 2013 using PowerPlex 21, some low‑level contaminant DNA was identified in the negative control. That was investigated at the time and assessed to be of sufficiently low level to be of no concern, and the results were passed for use. On further investigation in 2016, it was determined by PathWest that an insufficiently conservative approach had been taken in 2013. There was also a concern in 2016 that the scientist who had conducted the analysis in 2013 had not been notified of the contamination found in the negative control. The scientist who conducted the review in 2016 reviewed the results and noted that the contamination had been detected even after the batch in question was re‑run at the electrophoresis stage. In light of this a decision was made to withdraw the report in respect of the window frame. This also resulted in a comprehensive review of all of the negative controls in this case. That review did not identify any issues with the other negative controls.[42] Dr McDonald expressed concern with the handling of the investigation into this contamination event.[43]
Issue 1 - contamination - the wallet sample
[42] ts 198 - 199.
[43] ts 439.
In regard to the wallet sample, Mr Egan's evidence was that no contamination was found using Profiler Plus in 2009. However, when PowerPlex 21 retesting was done in 2013, some contamination was found in another negative control that was included in the same batch as the negative control for the wallet. The contaminated control was the same one (4422) referred to in respect of the bedsheet, that is, the control relating to the window frame. However, the wallet control was not itself found to be contaminated and, accordingly, the results were passed as valid to report.[44]
[44] ts 192, 279.
When the wallet sample was sent to Cellmark in 2016 for further testing, there was an approximately 30% increase in the quantity of DNA found in the sample. It was suggested by the defence that this was due to contamination. Mr Egan said that that was one explanation but that others included the removal of inhibiters that would release more available DNA or the use of a more sensitive technique to quantify the DNA.[45] Dr Walbank said that it was incorrect to compare the quantitation figures for PathWest with those of Cellmark because they used different tests. He said that it was possible that the Cellmark test had greater sensitivity. He said that the quantity of DNA was within expectations and he rejected the suggestion that the increase could be a result of contamination.[46] Further, Mr Turk said that, at low levels, the quantitation amounts may not be completely accurate.[47]
[45] ts 232 - 233.
[46] ts 378 - 379.
[47] ts 379 - 380.
Dr McDonald suggested that the increase in DNA could be explained by contamination. He said that a reduction of DNA would be expected over time, not an increase. He accepted that different procedures could produce different results, but not of this degree.[48]
[48] ts 344 - 345.
Both the original sample and the PathWest negative control were sent to Cellmark in 2016 for further testing. Cellmark used a different kit, NGM Select. Trace DNA was detected at one locus in the negative control by Cellmark. Dr Walbank said that this allele was also present in the original sample, but at a much higher level and was consistent with the reference profile of the deceased. He concluded that this contamination, even assuming it was also present in the original sample, would not have an impact on his interpretation of the results.[49] He said that it could not be assumed that contamination had occurred at PathWest.[50] There was no gross contamination that would give concern as to the reliability of the results.[51]
Issue 1 - contamination - the doorknob
[49] ts 284 - 285.
[50] ts 385.
[51] ts 386.
Mr Egan gave evidence that low‑level DNA was detected by PathWest in the negative control for the doorknob sample. The initial testing using Profiler Plus did not disclose any contamination in the control. However, in 2013, the negative control was re‑analysed using the PowerPlex 21 kit. This kit is known to have greater sensitivity. Four alleles were then detected in the control. Three of the four contaminant alleles were at sites only tested by PowerPlex 21, which would explain why they were not detected in the earlier testing (assuming that they were present at that time).[52]
[52] ts 195.
An investigation was conducted to compare the characteristics of the contaminant DNA in the control sample with the doorknob sample. When compared it was found that the characteristics of the contaminant DNA were not found in the doorknob sample. This means that the peaks detected in the control did not appear in the sample. This supports a conclusion that the contaminant in the control was not present in the sample. For that reason, the contamination of the control was not considered to be an issue and the sample was passed as valid for use.[53]
[53] ts 194.
One of the PathWest negative controls for the doorknob was sent to Cellmark in 2016 for further testing and analysis. Again, this testing was undertaken with a different kit, NGM Select. Three single alleles were detected at trace levels in that control. One of those alleles was an X component at the sex determinant locus. Dr Walbank said that an X component in the negative control would not have a particular bearing on the interpretation (since both males and females have an X). Of the other two alleles, one was not detected in the sample and could be discounted. The remaining allele was present in the sample, but at a higher level.[54] Mr Turk said that, because of this, he treated the equivalent peak in the sample as uncertain. This means that the existence of that peak in the sample is given a lower weighting in calculations. This is a way of allowing for the possibility of contamination.[55]
[54] ts 285 - 286.
[55] ts 287.
One of the alleles detected by Cellmark was at a site tested by PathWest, though PathWest did not detect that allele. Mr Egan said that there were several possible explanations for this. Different testing kits had been used and it was possible that Cellmark used a more sensitive technique. It is also possible that this was a different contaminant, not present during the PathWest testing. Further, it could be a stochastic effect that had not occurred earlier.[56] Dr Walbank said that, because the contaminants were present only at trace levels, he would not necessarily expect them to be detected in every analysis attempt.[57]
[56] ts 196.
[57] ts 287.
Cellmark also created a new negative control for the testing of the doorknob sample. A single allele was detected at a trace level in that control. Dr Walbank said that that allele was not present in the sample and thus did not affect the results. He said that a single component like this could be explained as a small extraneous piece of DNA that is in the environment of the laboratory. It is not particularly unusual to detect such trace levels of DNA in negative controls, given the sensitivity of the profiling kits that are used. He said such results served the purpose of showing that there was no gross contamination of the samples.[58]
[58] ts 287 - 288.
Mr Egan was also asked about inconsistent results in regard to the presence of amelogenin (the Y chromosome indicating a male contributor) in the door knob sample. In the original sample a Y chromosome was not detected, but after treatment with a chemical the presence of a Y chromosome was detectable, though at below the reportable threshold. Mr Egan accepted that it was possible that the Y chromosome was not present, but there were other possibilities.[59] Whether the contributor in question was male is open to question.[60] Dr McDonald was of the view that this inconsistency should not have occurred and was significant.[61] He suggested that the Y could be an artefact or drop in.[62] This issue is further considered in the section below dealing with contributor numbers in the doorknob sample.
Issue 1 ‑ contamination ‑ the thong
[59] ts 254.
[60] ts 258.
[61] ts 349 - 352.
[62] ts 356.
In relation to the thong, Mr Egan said that no contamination was detected by PathWest at any stage.[63]
[63] ts 196.
The thong sample and the associated negative control were sent to Cellmark for further testing. Again, the NGM Select kit was used by Cellmark. No contamination was detected in the PathWest samples.[64]
[64] ts 289.
Cellmark created its own negative controls. The thong sample was in the same batch as the Cellmark negative control for the doorknob that has already been referred to. Dr Walbank said that the single allele at trace level detected in that control was not present in the thong sample and so it was disregarded[65]
[65] ts 289 and see also ts 500.
Dr McDonald said that his reservations in respect of this sample were that there was a very small quantity of DNA and contamination in a negative control. He said that this was coupled with a low likelihood ratio, an indeterminate number of minor contributors and no indication of the false positive rate.[66]
Issue 2 - contributor numbers - general
[66] ts 358.
Dr Taylor accepted that opinions about contributor numbers can differ between scientists acting in good faith. He also said that while it is ideal to get the number of contributors correct, a number of tests had been undertaken to determine how much effect an overestimation or underestimation of contributors has on the likelihood ratios.[67] Whilst there is a degree of subjectivity in assigning contributor numbers, the risk of error is reduced by ensuring that every assessment is peer reviewed.[68]
[67] ts 142.
[68] ts 143.
It was put to Dr Taylor that if the estimated number of contributors was incorrect the likelihood ratio may also be affected. Dr Taylor said that this was not always the case. If there is a strong donor to the sample an increase in the number of contributors has no real effect on the likelihood ratio. If the queried person is only represented as a possible minor contributor then an increase in the number of contributors can decrease the likelihood ratio. If the number of contributors is underestimated, then known contributors could be falsely excluded.[69]
[69] ts 143.
Dr Taylor is the joint author of a paper dealing with the assessment of contributor numbers.[70] He accepted that one of the findings of that paper was that scientists differed as to the number of contributors in respect of a sample that had lower levels of DNA. He said that there is more complexity in assigning contributor numbers in such cases because the effect of stochastic events is more difficult to determine. Where the majority of the peaks representing alleles are high, it may be possible to identify low peaks as stochastic effects or artefacts. Where all of the peaks are low there may a difference of opinion as to whether a peak represents a contributor or is a stochastic effect. This means that there is a degree of subjectivity in the assignment of contributor numbers in respect of low level samples and scientists rely on their own experience and other factors.[71] In re-examination Dr Taylor expanded on this by saying that a scientist would draw upon knowledge of DNA profile generation, how the process works, how degradation of DNA may manifest itself and how the profile kit and the instruments in the laboratory operate.[72]
[70] Cooper et al, 'Investigating a Common Approach to DNA Profile Interpretation Using Probabilistic Software', Forensic Science International: Genetics, Vol 16 (2015), pages 121 to 131 - exhibit P1 page 291.
[71] ts 160 - 161.
[72] ts 184.
Dr Taylor was also asked about another paper of which he is a joint author.[73] One of the issues considered in that paper was the effect of an incorrect assignment of the number of contributors in a profile. The authors make the point that the number of contributors to a mixed profile is never known with absolute certainty. Assigning the number of contributors in a low level sample is more complicated, particularly where there are peaks close to the limit of detection or additional peaks just below the analytical threshold. This may result in the examining scientist assigning an additional contributor. However, it is important to clearly understand the effect of over or under estimating the contributor numbers:
Overestimating the number of contributors to a profile has the potential to generate an LR that favours the inclusion of known non-contributors, whereas underestimating the number of contributors has the potential to generate an LR that favours exclusion of a known contributor (page 103).
[73] Bright et al, 'Searching Mixed DNA Profiles Directly Against Profile Databases', Forensic Science International: Genetics, Vol 9 (2014), pages 102 to 110 ‑ exhibit P1, page 260.
Dr Taylor was asked about an experiment referred to in the second paper. The suggestion was that the experiment supported a suggestion that an error in the assignment of contributor numbers might result in a likelihood ratio that falsely favoured a person being a contributor. The experiment in question (experiment 3) involved creating eight low level samples that appeared to have two contributors based on allele count, though they in fact had three. These samples were then interpreted assuming both two and three contributors. Fewer adventitious matches were observed when an assumption of two contributors was made. Where three contributors were assumed, the number of adventitious matches rose from 1.5% to 86.3% of the database size. Almost all of these matches were at the very lowest level of probability.[74] Defence counsel relied on this as showing that an erroneous overestimation of contributor numbers could produce a result that falsely included a person. However, as Dr Taylor explained, the high number of low‑level false inclusions occurred because a third person was not required in order to explain the observable profile. By assuming a third contributor, the STRmix program would assume that alleles had dropped out at every site that was examined.[75] This means that the third contributor could match a large number of possible contributors in the population, which accounts for the very low level of probability for those adventitious matches. I note that this detracts from the defence argument because it justifies assigning the least number of contributors that can explain a profile in order to minimise adventitious matches. It also tends to show that even if the number of contributors is overestimated, the risk of producing an adventitious match with a high likelihood ratio (like those produced in this case) is small. The risk associated with underestimating the number of contributors is that a known contributor will be falsely excluded: which operates to the advantage of an accused person.
[74] Bright et al, 107, see also ts 458 - 459.
[75] ts 164.
Mr Egan was also asked about this paper and experiment. In his statement he noted that the adventitious match with the highest likelihood ratio in experiment 3 had been produced without applying correction factors that are routinely used at PathWest. When those factors were applied, that ratio reduced to a much lower level. He said that this experiment demonstrates the danger inherent in overinflating the number of contributors in a mixed DNA profile when the data does not support such a conclusion.[76]
[76] Statement of Mr Egan, exhibit P1, page 150.
Mr Egan accepted that there is a risk of adventitious matches where the likelihood ratios are low. In particular, he said that where the ratio was 100,000 or less in favour of inclusion, there was potential for the person in question to in fact not be a contributor and for their profile elements to be present by chance. He said that PathWest uses a conservative cut-off of a million and for anything under that figure the possibility of adventitious matching has to be considered.[77]
[77] ts 242 - 243.
Mr Egan was asked about another paper.[78] He accepted that this study tended to indicate that as contributor numbers get beyond three the ability to accurately determine contributor numbers declines. However, he said that the additional information available in some kits reduces the probability of incorrectly identifying the number of contributors. He said that the kit now used by PathWest has got 'a lot of information in it'.[79] I also note that the paper states that calculations were based on allele frequencies only and that peak heights were ignored. Dr Taylor referred to this in his evidence, stating that in practice analysts always take peak heights into account when assessing contributor numbers.[80]
[78] Coble et al, 'Uncertainty in the Number of Contributors in the Proposed New CODIS Set', Forensic Science International: Genetics, Vol 19 (2015), pages 207 ‑ 211 - exhibit D5
[79] ts 248.
[80] ts 165.
A fourth paper put to Mr Egan raised the same issue.[81] This paper considered the effect of masking, where two people share the same allele at a site. The authors state that there is an increasing risk of incorrectly inferring the number of contributors as that number increases. That risk can be reduced by using a kit that tests more sites and is therefore more discriminating. The use of quantitative information such as peak heights and allele frequencies can also ameliorate the risk, but not remove it entirely. The authors conclude that the best approach for dealing with uncertainty as to contributor numbers is to incorporate it into the calculation of the likelihood ratio. Mr Egan said the paper was highlighting issues that were well understood by scientists analysing mixtures.[82]
[81] Curran and Buckleton, 'Uncertainty in the Number of Contributors for the European Standard Set of Loci', Forensic Science International: Genetics, Vol 11 (2014), pages 205 ‑ 206 - exhibit D10.
[82] ts 250.
Dr Taylor explained that the rule of assigning the least number of contributors required to explain a profile is used because it is a conservative method of interpretation. If more contributors are assigned than is required, the risk of falsely including people grows. He said that 'we would rather falsely exclude people than falsely include people to profiles and so we choose that minimum number that can explain what we seek'.[83]
[83] ts 182.
It was put to Dr Taylor that it becomes more difficult to make an accurate assessment of contributor numbers once the possible number is beyond three. He agreed that there was an increasing chance that a higher order mixture would masquerade as a lower order mixture, that is, that a mixed sample may appear to have less contributors than in fact it does (as was the case with experiment 3). But he did not view this as a problem in terms of producing reliable results. He illustrated this by pointing out that whilst it is usual to assign the same number of contributors when comparing the prosecution and defence hypotheses, it was open to the defence to choose their own number. But increasing the number of possible contributors beyond what is necessary to explain the profile is to the detriment of the accused because it would have the effect of increasing the likelihood ratio in favour of the prosecution hypothesis.[84]
[84] ts 167.
It was put to Dr Taylor that he had not complied with quality assurance standards. The relevant standard is ISO/IEC 17025, which has been adopted by NATA. NATA has published an 'Application Document' that provides interpretative criteria and recommendations for the application of ISO 17025 for accredited facilities.[85] Clause 5.4.6 of the Application Document states that an estimation of uncertainty of measurement is required for all quantitative tests. It was put to Dr Taylor that the assessment of contributor numbers was a quantitative test and that he had not provided an estimation of the uncertainty involved in that assessment. He said that the effect differences in contributor numbers may have is well known, but to quote a specific level of uncertainty for each unique profile would be very difficult. He said that his report did state that the most likely number of people that would be required to generate a profile is the number used, whilst acknowledging that it is always possible that the observed DNA profile is from a greater number of people. He said that he cannot put a numerical value on his degree of uncertainty in contributor numbers.[86] In re-examination he said that in the event it was not possible to make an accurate estimate of the number of contributors for a particular sample it would not be further analysed.[87]
[85] Exhibit P1, pages 542 - 565.
[86] ts 170 - 173.
[87] ts 186.
Mr Egan was also asked about the application of cl 5.4.6 to the estimation of contributor numbers. He said that he was not aware of any method for estimating the uncertainty of a judgment as to contributor numbers. The relevant clause also states that 'For qualitative analysis, facilities are encouraged to have an understanding of the variability of their results where this is possible'. Mr Egan said that PathWest staff members do have an understanding of the limitations of the techniques used in the laboratory, an understanding that estimations of contributors at low levels are more difficult and an understanding of DNA behaviour and stochastic effects. There is also an understanding of the statistical ramifications of the conclusions that are drawn.[88]
[88] ts 268 - 269.
Mr Turk said that whilst the number of contributors is a factor relating to likelihood ratios, the effect of an increase in possible numbers is not uniform. The effect will depend on the overall level of the profile and the level at which the person of interest is represented. It is not to the advantage of an accused person to postulate an increased number of contributors. He said that, in simple terms, the number of detected alleles does not increase, thus an increase in assumed contributors effectively means that those alleles are shared amongst a greater number of people. The effect of this is to produce a likelihood ratio in favour of inclusion that is similar to, or even slightly higher, than the original calculation.[89]
[89] ts 290 ‑ 291, see also ts 478.
Dr Walbank said that the assessment of contributor numbers in a mixed sample can be complicated by artefacts of the amplification process. This is the process whereby the DNA is replicated to increase the sample size. One of the known artefacts of this process is a small peak either before or after a detected peak. This is known as stutter. Dr Walbank said that stutter had been specifically investigated as part of Cellmark's validation of the NGM Select kit. This process identified expected stutter levels at each of the loci tested by that kit. This enables scientists to apply a judgment as to whether a peak is a stutter artefact or an allele.[90]
[90] ts 292.
Mr Turk said that the number of contributors can never be known with 100% certainty, but that a peer review process ensures that every result is verified by two or more scientists before it is reported.[91] This is intended to minimise any subjectivity in the assessment.[92] Any uncertainty as to the number of contributors is taken into account, not by giving a numerical figure for the degree of uncertainty, but by running the calculation several times to ensure it is repeatable, accepting that it is only an approximation, and only reporting the lowest of the various calculations undertaken and rounding down in favour of the defence position.[93]
[91] ts 293.
[92] ts 371.
[93] ts 372.
Dr Walbank said that the assessment of contributor numbers is both a general skill and one that also requires familiarity with the specific test kit that is used, particularly when dealing with low level complex results. The NGM Select kit includes one locus (SE33) that has been chosen specifically because of the large variability that occurs in the population at that locus. This means that results at that locus are particularly informative when making decisions about the possible number of contributors.[94]
[94] ts 298 - 299 and see also ts 481.
It was put to Dr Walbank in cross‑examination that Cellmark had declined to provide the defence with access to the results of its proficiency testing. He denied this and said that if specific tests relevant to this matter are requested Cellmark is happy to provide access. Whilst he was not closely involved with management of the proficiency testing process, he was not aware of any tests that related to low level samples with the same number of contributors as the samples in this case.[95] Mr Turk said that the accreditation process does involve review of file work involving complex mixed samples.[96]
[95] ts 374 - 375.
[96] ts 375 - 376 and see ts 483 - 486.
Dr McDonald said that an assessment of contributor numbers will have little effect for an identifiable major contributor to a mixture, but it will have an impact on minor contributors. The attribution of too few contributors will artificially exclude some minor contributors. The more contributors the more risk there is of contributors sharing alleles. At low levels there is a risk of drop out or stochastic effects. These factors can compromise the accuracy of an assessment of contributor numbers.[97]
[97] ts 316 - 318.
Dr McDonald said that in Europe a different approach had been taken. It involves taking into account how common types of alleles are at a particular locus. Some alleles are very common in the population and others are very rare. Information regarding the incidence of a particular allele is taken into account in assessing contributor numbers. This was contrasted with what was called the 'allele count' method, as used in this case. Dr McDonald referred to a study that showed that both methods were accurate where there were two or three contributors, but that the European method was more accurate as the number of contributors increased to four or five.[98] This was relied on as showing that where the number of contributors is more than three there will be problems in getting the number right.[99] I note that this paper was not put to the other witnesses for comment, that the allele count method proved to be at least (if not more) accurate for two and three person mixtures, that the accuracy of the allele count method improved as more loci were tested, and that the paper only considered up to 14 loci (plus amelogenin), being less than is examined in both PowerPlex 21 and NGM Plus.
[98] Haned et al, 'Estimating the Number of Contributors to Forensic DNA Mixtures: Does Maximum Likelihood Perform Better than Maximum Allele Count', Journal of Forensic Sciences, (2011) Vol 56 No 1, pages 23 - 28 ‑ exhibit D11.
[99] ts 322 - 325.
Dr McDonald suggested that if the number of contributors was increased for any of the samples, the likelihood ratio in respect of minor contributors would tend to go down, but said that he would prefer to avoid any prediction. He agreed that such calculations could be done, but said that he could not do them because he does not use either the STRmix or Like LTD software. He could not speculate on the possible results of any such calculations.[100] He also accepted that the risk of adventitious matches increases if the contributor number is increased.[101]
[100] ts 325 - 326.
[101] ts 326 - 327.
Dr McDonald suggested that the profiles obtained could be compared to the whole available DNA database to determine how many false positives were produced. It was not suggested that this had ever been done or how it could be done. He said that the risk of adventitious matches increases as the complexity of a mixture increases. He expressed the view that the risk is high where the likelihood ratio is below one million in favour of inclusion. However, he also accepted that the figure of one million was a conservative figure that had been set because it could be comfortably said that adventitious matches above that figure were extremely unlikely.[102] He could not point to any evidence that would establish conclusively that the assessments of contributor numbers by the prosecution experts were wrong.[103]
Issue 2 - contributor numbers - the bedsheet sample
[102] s 329 ‑ 330, see also cross‑examination ts 460.
[103] ts 462.
The bedsheet sample, when initially tested using Profiler Plus, was assessed as being consistent with being a mixed DNA sample from two people. The deceased and the accused could not be excluded as possible contributors but PathWest did not have the ability to do a statistical analysis for a mixed sample generated by the Profiler Plus kit. They were accredited to use the STRmix program for PowerPlex 21, but the PowerPlex 21 results were such that it was not possible to determine how many people were in the mixture. It was for this reason that the profile obtained from Profiler Plus for the bedsheet was sent to FSSA. That laboratory had performed the validation procedures for analysing Profiler Plus results with STRmix.
When asked why it was possible for PathWest to determine the number of contributors from the Profiler Plus results but not those from PowerPlex 21, Mr Egan said:
The ability of PowerPlex 21 to obtain more DNA profile is well known. It's reported as being approximately 30% more sensitive than Profiler Plus. So potentially there's more DNA in the system using that PowerPlex 21 DNA profiling system. Additionally to that, other DNA could be detected that wasn't present in the Profiler Plus sample, which adds some complexity to the interpretation. There's still DNA present, but because of our reporting guidelines, we were unable to tell the most accurate amount of contributors that make up that DNA profile. And if we can't accurately determine the number of contributors then we can't analyse it further.
HALL J: Does that in any way suggest that the Profiler Plus results were unreliable, if PowerPlex is showing up things that perhaps have not been detected by Profiler Plus?‑‑‑No. That's exactly how the system should work. If you've got a technique that's more sensitive, you will get more information with a more sensitive system. And a lot of the samples that we produce in our laboratory from Profiler Plus then upgrade to PP 21 do show significant improvement. Sometimes they don't show improvement to the point that we can use them. And this is an example of one of those.
All right, but you're saying the fact that you subsequently got results with PowerPlex with an undetermined number of contributors does not in any way reflect upon the validity of the Profiler Plus results?‑‑‑Not in any way, no (ts 209 ‑ 210, see also 275 ‑ 278, 405 and 418).
Mr Egan said that the assessment of contributor numbers is based on both experience and the academic literature. This involves not only counting peaks, but also taking into account the height of peaks. In relation to the bedsheet sample, Mr Egan said that he was 'quite confident' that it is a two person mixture. He said there was nothing in the electropherogram to suggest that there were any more than two contributors to this sample.[104] Mr Egan also said that PathWest routinely used peer review and that if there was no concordance between two scientists as to the minimum number of contributors, or there was doubt as to the number, then the profile would be discarded.[105]
[104] ts 211 - 212.
[105] ts 240, 243, 261.
Dr Taylor said that on receiving the Profiler Plus results for the bedsheet he made an independent assessment of the number of contributors. His view was that there were two contributors to that mixed sample. He said that when a scientist assigns a number of contributors it is on the basis of the minimum number of contributors that can reasonably explain the profile. He was 'quite confident' that two people could explain this particular DNA profile.[106]
[106] ts 138 - 139.
Dr Taylor was asked for his view as to why it was possible to determine the number of contributors using the Profiler Plus results but not with PowerPlex 21 results. He could only comment generally because he had not seen the PowerPlex 21 results, but said:
PowerPlex 21 is a more sensitive kit, so it's liable to detect and amplify more information than Profiler Plus. Now that additional information could confound someone's interpretation of the number of contributors, meaning that they won't go on to analyse it in a program like STRmix. The other possibility could be that if there has been some time elapsed between the initial analysis with Profile Plus and the subsequent analysis in PowerPlex 21 then the DNA could have degraded during that time, which could also complicate the interpretation of the profile produced by the second PowerPlex 21 analysis (ts 139 ‑ 140).
In cross‑examination of Mr Egan it was put to him that in 2013 he had done a STRmix calculation on the PowerPlex 21 sample on the basis of two possible contributors, which had excluded the accused as a contributor. He agreed that he had done this but that on reviewing the sample it was not possible to say with any degree of confidence how many contributors there were. For this reason the tentative exclusion result was not suitable for reporting because it was not an accurate reflection of the available data.[107] I note that the false exclusion of known contributors is a recognised consequence of underestimating the number of contributors. Dr McDonald relied on this result as being inconsistent with the analysis later undertaken by Dr Taylor.[108]
[107] ts 420.
[108] ts 359.
Dr McDonald was of the view that the PowerPlex 21 results were not consistent with the accused being a contributor and the fact that these results were not provided to Dr Taylor was significant because of the difference between them. He said that 'some sort of investigation' was required.[109]
[109] ts 426 - 427.
Dr McDonald also pointed to there being more information contained in a post- clean-up Profiler Plus result (exhibit D13) than in the pre-clean-up result (exhibit D12). The 'clean-up' referred to is the use of a chemical to remove factors that can inhibit the expression or amplification of the DNA (often called Qiagen clean-up). However, he accepted that it would be expected that more information would become available as a result of the clean-up procedure and that was a possible explanation for that difference.[110]
[110] ts 429.
Dr McDonald said that there were peaks detected in the Profiler Plus result (exhibit D13) that were not present in the PowerPlex 21 result (exhibit D3). However, when the relevant electropherograms were compared it was apparent that the differences relied on arose from differences in peak heights and not on the total absence of peaks. Some of the peaks in the PowerPlex 21 result were present but at below reportable heights.[111] Nevertheless, Dr McDonald maintained that the PowerPlex 21 result contained no information that supported the accused being a contributor.[112] However, he also accepted that there was nothing in that result that excluded the accused.[113] He said that an investigation into a possible contamination event needed to be conducted.[114]
Issue 2 - contributor numbers - the wallet sample
[111] ts 430 - 431.
[112] ts 431.
[113] ts 433.
[114] ts 434.
Dr Walbank concluded that there were indications of at least four, and possibly at least five contributors, to the wallet sample. The possibility of five contributors arose only because at one locus there was an indication, at a trace level, of an additional contributor. Mr Turk came to the same conclusion.[115]
[115] ts 295 ‑ 296 and see 486 - 487.
It was put to Dr Walbank in cross‑examination that the low quantity of DNA and the fact that one of the contributors accounted for 90% of the DNA meant that there was a very high risk of miscalculation of contributor numbers. He did not agree with that suggestion, rather he described the risk as a possibility. That possibility was accommodated by incorporating the possibility of five contributors into the calculations.[116]
[116] ts 487 - 489.
Dr McDonald took issue with the assumption that the deceased was one of the contributors. This was despite the fact that the profile of the contributor present at a higher level matched that of the deceased. He said that he would like calculations to be done on the basis that the deceased was not a contributor, as well as on the basis that he was. He said that if the deceased was not a contributor the numbers would become too complicated for a likelihood calculation.[117] In his view there is a high risk of underestimating the contributor numbers with respect to this sample.
Issue 2 - contributor numbers - the doorknob sample
[117] ts 347.
Dr Walbank concluded that there were indications of at least three contributors to the doorknob sample and that at least one male contributed to that result. Mr Turk was of the view that there were at least two, or possibly at least three contributors; there being only a relatively small indication of a third contributor. He did his calculations on the basis of a three person mixture to take into account the indication of a third contributor. He did two calculations - one on the basis that the indication of a third person was a drop‑in at a single locus (an extraneous environmental contamination) and one on the basis that there were three complete contributors. He reported the likelihood ratio that was the lowest in favour of inclusion, that is, the one most favourable to the accused.[118]
[118] ts 297 ‑ 298 and see 491 - 495.
In cross‑examination it was put that there were inconsistent results in the three tests conducted as to whether the Y chromosome was present, and that this meant that it could not safely be said that there was a male contributor. Dr Walbank said that the absence of a Y marker cannot be taken to mean that there was no male contributor as the Y chromosome can drop out in the same way as any other allele. In his experience, other NGM Select results have shown an absence of a Y marker but when re-tested using a test specifically designed to detect that chromosome it has been shown to be present. That test was not considered necessary in this case because the NGM Select kit had performed within expectations.[119]
Issue 2 - contributor numbers - the thong sample
[119] ts 495 - 497.
Dr Walbank concluded that there were indications of at least three contributors to the thong sample and at least one male had contributed to that result. There were no indications of any more than three contributors at any of the sites tested. Mr Turk came to the same conclusion.[120]
Issue 3 - proficiency and validation - PathWest
[120] ts 296 ‑ 297 and see 498 - 499.
There was a suggestion by the defence that the practices and procedures at PathWest were deficient and that contamination events in the past cast doubt on the reliability of its testing in general. I understood this to also be relevant to the analysis undertaken by FSSA and the further testing and analysis undertaken by Cellmark because the samples considered by each of those laboratories originated from PathWest.
In his statement Mr Egan said that the PathWest laboratory is accredited by NATA for forensic biology techniques including DNA analysis. It operates under the Australian Standard ISO 17025:2005 General requirements for the competence of testing and calibration laboratories. Accreditation assessment is performed by qualified independent experts. The parameters of accreditation include organisation and management, methods and procedures, quality assurance, equipment, reports, procurement of services and supplies, and accommodation and safety. All staff are formally trained in the methods and techniques they are required to employ. The laboratory participates in regular external and internal proficiency tests. There are regular audits with both internal and external auditors.[121]
[121] Exhibit P1, page 132.
Mr Egan said that PathWest receives approximately 10,000 to 12,000 cases per year which equates to between 45,000 and 50,000 samples for testing. There has been a 5% to 10% increase in tests per year between 2009 and 2013. As to suggestions that there was a systemic failure of the required quality controls by PathWest, Mr Egan said that the quality systems at PathWest had been passed for accreditation purposes since 2001. He said that, as with any human process, there will be mistakes but that the purpose of a quality system was to pick those up. NATA has reviewed PathWest in depth over the last 16 years and passed it as suitable for reporting DNA results. Whether there have been contamination events and the processes for dealing with them are matters which NATA looks at in the quality assessment each year.[122]
[122] ts 203 - 204.
It was put to Mr Egan that PathWest had not validated the reliability of testing for low level mixed samples of DNA. He denied this and said that when PowerPlex 21 was introduced at PathWest part of the validation process was to do low level tests of known samples to make sure the kit was able to detect down to a certain level and that STRmix could then be used to produce an accurate result.[123] That validation had been passed by NATA.[124]
[123] ts 264 - 265.
[124] ts 266.
To the suggestion that a scenario precisely like each of the samples in this case had not been validated, Mr Egan said sufficient validation had been done to give confidence as to the reliability of the procedures at low levels and that it was not reasonable to 'validate every single scenario or else we would be validating our whole lives and not actually doing any real work'.[125] Equally, Mr Egan said that proficiency testing of scientists was undertaken on a regular basis, but he was not aware of tests that specifically related to very low level complex samples.[126]
[125] ts 265.
[126] ts 266.
The fact that PathWest has been accredited by NATA, that it has been regularly reviewed, that the tests it uses have been validated and that the scientists involved have undertaken proficiency testing was not the subject of any evidence controverting that given by Mr Egan. Whilst his evidence was challenged, he maintained his position in cross-examination and there was nothing inherently unlikely or contradictory about his evidence. To the extent that it was suggested that PathWest had not complied with the relevant standards in respect of the analysis of complex mixed samples, no evidence was led from any witness from NATA that the standards and procedures used by PathWest were deficient in this regard.
Dr McDonald said that he had been involved in the implementation of ISO 17025 for DNA testing in Australia in 1996, and for some years conducted inspections as an accreditor and auditor of laboratories conducting paternity tests.[127] He had last done work as an assessor or auditor in 2003 and last worked in an accredited laboratory in 2004.[128] He said that he only had a copy of his accreditation as an assessor of medical laboratories; he said NATA appeared to have lost his records in respect of forensic laboratories. His training as an assessor was a one‑day session. He is no longer on NATA's list as an assessor of forensic laboratories. He has not attended any training sessions by NATA since 2003.[129]
[127] ts 331, 464.
[128] ts 440, 446.
[129] ts 464 - 466.
Dr McDonald said that the Standard required a laboratory to show that it could conduct tests on known samples and produce reliable results. He said that the samples should be representative of those that are tested in the laboratory. He said that he was unable to find any published validation studies for PathWest in respect of samples like those tested in this case. He said that the doorknob sample was at a lower level than any he had seen in any validation study.[130] In his view, the quantities of DNA in this case were at such low levels and of such complexity that no valid testing could be done.[131] He later said that each of the laboratories had conducted validation studies but that those studies had confirmed that the number of contributors in low level samples cannot be reliably determined.[132] He also accepted that his opinions were based on his interpretation of the Standard.[133]
[130] ts 332 - 333.
[131] ts 334.
[132] ts 425 ‑ 426, 442.
[133] ts 467.
Dr McDonald said that the proficiency testing undertaken by an independent external body (Collaborative Testing Services) includes a wide range of different types of samples, but not low level DNA samples. There can be mixed samples, but they are always at optimal levels. He was aware of a study in the United States in which low level samples had been sent out and the range of responses was 'enormous' and 'the possibility of standardization was virtually non‑existent'.[134]
[134] ts 334 - 335.
Dr McDonald was of the view that ISO 17025 requires the degree of uncertainty as to the assessment of contributor numbers to be reported. He suggested that this could be expressed as a degree of certainty and that it was measurable. He said that this was demonstrated by the papers referred to (those authored by Haned, Coble and Curran). He said that the reports in this case had failed in this respect.[135]
Issue 3 - proficiency and validation - FSSA
[135] ts 336 - 337.
Dr Taylor was also asked whether there were any validation studies dealing with the interpretation of low level mixed samples. He referred to two studies, one authored by himself published in 2014 and another conducted by the FBI and published in Forensic Science International: Genetics. He also referred to two other papers dealing with the estimation of contributor numbers.[136]
[136] ts 179 - 180.
There were no questions specifically challenging the validity of the methods used by Dr Taylor or his proficiency in respect of them. There was an issue raised as to the significance of a STRmix calculation done by PathWest on the PowerPlex 21 results, which was inconsistent with the result produced by Dr Taylor. But that other result was not raised with Dr Taylor and Mr Egan explained why it was unreliable and not reported on.
Issue 3 - proficiency and validation - Cellmark
In regard to practices and procedures at Cellmark, Dr Walbank said that Cellmark is assessed on an annual basis by an external accreditation service.[137] Cellmark is a private scientific forensic services business with a little over 500 employees, of whom about 240 are scientists, at two sites in the UK. The majority of the work undertaken by Cellmark is DNA analysis. There are typically around 600 ongoing cases at any time, with 20 or 30 reports being delivered each day. The largest client base is the various police forces in England and Wales, but work is also done for other government agencies and for the defence in criminal cases. There are two other companies, one of similar size and one significantly smaller, that together with Cellmark provide the majority of forensic services in the UK and have taken over from the government forensic agency when it was closed down several years ago.[138]
[137] ts 289.
[138] ts 369 - 370.
Dr Walbank said that he has nearly 20 years' experience in interpreting DNA profiles and has received training in the use of NGM Select. Cellmark commenced using NGM Select in July 2014. All scientists at Cellmark take part in monthly competency testing in the interpretation of mixtures. He said that he had carried out hundreds, if not thousands, of assessments in respect of low level complex DNA mixtures.[139] In the case of each of the samples reviewed at Cellmark for this case, Dr Walbank made an initial assessment of contributor numbers and Mr Turk then made a separate, independent assessment.[140]
[139] ts 293 ‑ 294, 299.
[140] ts 294.
Mr Turk said he had been interpreting results since 1999 and had specialised in low template DNA since 2004. He estimated that he interpreted up to 10,000 complex low level results and this had been the bulk of his work load for many years. He too has undertaken competency testing in the use of NGM Select. Mr Turk accepted that the quantities involved in the samples he examined were small, but that the results had been obtained using routine profiling methods; specialist high sensitivity enhancement work had not been employed (so called phase 2 and 3 methods). In his view, the level of the DNA in these samples did not have any impact on the validity of the likelihood ratios.[141]
[141] ts 294 - 295.
Mr Turk said that before NGM Select and Like LTD were implemented they were subjected to validation by a research and development team at Cellmark. This involved running a large number of samples to ensure that the systems perform to expected standards. Operational validation was then conducted to establish how the systems operate in respect of case work.
Mr Turk explained that the name Like LTD is a contraction of the phrase 'likelihood of low template DNA'. It is a software program developed by Professor David Balding and a team at University College, London. It is a tool specifically designed to assist in evaluating likelihood ratios where small amounts, or complex mixtures, of DNA have been detected. It has been in use at Cellmark since 2014. Unlike earlier models which relied on manual calculations of well-amplified DNA, Like LTD is able to take into account factors such as degradation or drop out that may occur with older DNA. Results that are clearly alleles can be designated as certain and others that are low or possibly artefacts can be designated as uncertain. The peaks with an uncertain rating will be given a lower weighting in the probability calculation than those that are in the certain category.[142]
[142] ts 301 - 304.
Mr Turk said that Like LTD had been the subject of a number of peer-reviewed articles in prestigious journals, describing the mathematics and showing the functionality of the program. The software is also 'open source', meaning it is freely available for any person to use and to be scrutinised by other academics and institutions. A full user guide is available on a website associated with the program. Mr Turk has been using the program since 2014 and Cellmark was accredited to use it by the United Kingdom Accreditation Service (UKAS) (using ISO 17025) in October 2015. The training at Cellmark included seminars conducted by Professor Balding and his associates.[143] There are ongoing assessments by UKAS, the most recent of which was in April 2017, during which Mr Turk was required to demonstrate his use of Like LTD.[144]
[143] ts 304.
[144] ts 307.
Mr Turk said that Professor Balding had been using Like LTD to present findings at trials in the UK Crown Court since 2010. The program was refined and developed over the following years. The scientists from Cellmark have been presenting such evidence since 2014. Both Mr Turk and Mr Walbank have provided reports and given evidence in that regard, frequently without challenge.[145] Both NGM Select and Like LTD have achieved widespread acceptance in the scientific community in the UK.[146] I note that the use of Like LTD was referred to by the Scottish High Court of Judiciary in Her Majesty's Advocate v Sinclair.[147] Evidence relying on Like LTD was also led in the recent trial of State of WA v Wark [No 2],[148] though the admissibility of that evidence was not challenged.
[145] ts 308 - 309.
[146] ts 363 - 364.
[147] Her Majesty's Advocate v Sinclair [2014] HCJAC 131 at [114] ‑ [115].
[148] State of WA v Wark [No 2] [2018] WASC 18 [714] - [724].
In his witness statement, Mr Turk provided a list of the cases in which Professor Balding has provided evidence. In the majority of the cases, the evidence has been accepted without challenge. In four cases the evidence had been challenged, only one of which was initially successful (the evidence was admitted at a re-trial).[149] Mr Turk said that the case in which there had been the initial successful challenge, R v Dos Santos (Unreported, London Central Criminal Court, March 2012), related to one particular function of the program that was subsequently modified by Professor Balding, leading to the acceptance of the evidence at a re‑trial.[150]
[149] Statement of Mr Turk, exhibit P1, page 35.
[150] Statement of Mr Turk, exhibit P1, page 36.
Mr Turk said that on two occasions his evidence in respect of Like LTD had been challenged, once at a trial at the Old Bailey in London and once at the Maidstone Crown Court. On both occasions the challenge had been unsuccessful and the evidence had been accepted.[151]
[151] ts 364.
Mr Turk said that the results in this case were similar to others he had reported on. Given the sensitivity of the tests that are now used, he considered the results here to be quite routine. Samples with multiple contributors at low levels were a 'day‑to‑day finding'. The results here were typical of others that he had presented in courts. Dr Walbank said that much of his work in the last six years had centred on the use of high sensitivity techniques, which generate similar profiles to those in this case. It was not necessary to use that technique in this case because NGM Select is a comparatively sensitive and more robust test than some others.[152]
[152] ts 365 - 367.
Dr Taylor was also asked about Like LTD, though it was not the program he had used. He was asked whether that program was less reliable and less capable of producing accurate results than STRmix. He said that he did not agree with that proposition. He said that Like LTD is what is known as semi‑continuous software, rather than the fully continuous software used in STRmix. In practical terms, this means that STRmix includes information regarding peak heights in its calculations, whereas Like LTD does not. As a consequence, STRmix has more power to discriminate and will tend to produce larger likelihood ratios in favour of true contributors and larger ratios in favour of the exclusion of non-contributors for mixed samples that have divergent proportions. However, he did not accept the suggestion that results from STRmix were necessarily more accurate or reliable. He said that results from both programs were 'perfectly accurate and reliable and correct insofar as they correctly apply the biological and mathematical models that they use to evaluate profiles'. He said that the difference could be compared to that between the testing kits, Profiler Plus and PowerPlex 21; one simply uses more information than the other and is therefore more discriminating.[153]
[153] ts 168 - 169.
Dr McDonald said that he had seen a 'limited comparison' of STRmix and Like LTD when used on the same samples. He said that he was 'somewhat disturbed' by the variation in the results. He also said that as mixtures became more complex STRmix could provide a result whereas Like LTD could not. However, he conceded that he had no personal experience with either software program, and that he was not a mathematician and was unable to explain the value of evidence produced by one program as opposed to the other.[154]
[154] ts 337 - 338.
Dr McDonald did not dispute that Cellmark had been accredited and reviewed by the UK accreditation authority. Nor could he dispute the evidence that the analysis of samples like those in this case were routinely reported on by Cellmark. Nonetheless, he maintained that Cellmark had not validated their system sufficiently to justify the evidence given by Dr Walbank and Mr Turk.[155]
[155] ts 468.
An Alternative Suspect
Some questions were put to witnesses regarding the possibility that some of the alleles observed in the profiles were also consistent with the reference sample of another person. The suggestion appeared to be that, if likelihood ratios were calculated, they may indicate that the other person was a possible contributor to the profiles. This questioning did not advance beyond speculation and it was not submitted that this possibility was a ground for rejection of the prosecution evidence.
In any event, as the State pointed out in its submissions, the prosecution case does not exclude the possibility that others besides the accused were involved in the murder. Accordingly, evidence that may implicate other people does not necessarily exculpate the accused or diminish the significance of evidence that implicates him.
Conclusions
The focus of the application is on the admissibility of the evidence. As I have earlier noted, it is not necessary or appropriate for me to make any assessment of the credibility of the witnesses, the final conclusions that should be drawn from that evidence or the weight that should be given to the evidence by the jury (or finder of fact in the event that the accused applies for a trial by judge alone).
The question for determination is whether the evidence is the product of a scientific process that is sufficiently established and reliable as to make the opinions of the prosecution witnesses admissible at the trial. The fact that the evidence is disputed does not in itself mean that the evidence is inadmissible. Nor does the existence of different opinions of scientists in respect of the same matter necessarily mean that the evidence is unreliable. It is common for trials to involve competing expert opinions.
The contamination issues in respect of the samples have been properly investigated and taken into account. I am not convinced that the contamination that was detected leads to a conclusion that the results are invalid or such that no reliance could reasonably be placed upon them. The fact that there was some contamination detected and the methodology for dealing with it are matters that the jury can properly take into account in assessing what weight they give to the evidence. It would be open to a jury to conclude that the evidence is reliable notwithstanding the contamination. The difference of views between the prosecution witnesses and Dr McDonald is also a matter for the jury.
The contributor number issue is essentially one of differing opinions by scientists. The prosecution witnesses accept that this involves an exercise of judgment; a judgment that is not arbitrary but based on experience and knowledge. To the extent that a different number of contributors has been suggested as possible, some alternative calculations have been done. It would be open to the defence to lead other calculations if they choose to do so. But that does not detract from the evidence of the prosecution experts. I am satisfied that the assessment of contributor numbers was properly done and was in accordance with accepted principles and practices. To the extent that Dr McDonald has a different view as to the numbers, or whether it is possible to make any assessment at all on the available evidence, that is a dispute which is appropriate for the jury to resolve.
The issue of validity of the methodology has two components. As to whether Like LTD is reliable, I accept that it is a probabilistic program, similar to STRmix. The available evidence is to the effect that it is reliable and is an extension of DNA science, rather than being some new or discrete area of knowledge. There was no exploration of the underlying mathematical basis of the program and no evidence led to controvert the statements of the prosecution witnesses that the program has been validated and is now widely accepted in the UK. As to whether each of the laboratories and scientists has proven experience in testing low-level samples, they each gave evidence that they do. The challenge to that evidence by the defence is a matter that goes to weight and is for the jury to determine.
As to whether the evidence should be excluded in the exercise of the discretion to exclude evidence where its probative value is outweighed by its prejudicial effect, the submission on behalf of the accused essentially relies on the claim that the jury would be incapable of making an assessment of the evidence. The suggestion is that the jury would be over-awed by the statistical results and accept the evidence without considering the issues regarding the reliability of the testing and the methodology used. I do not accept that suggestion. Any jury will readily understand that the statistical calculations are dependent on the reliability of the tests and on the accuracy of the judgment of the scientists involved. It will be apparent that the defence expert has a different view in a number of critical respects and that these are issues that the jury will have to resolve by carefully considering all of the relevant evidence and the credibility of the witnesses. Directions by the trial judge would ensure that the jury understand the nature of their task.
For these reasons, I ruled that the evidence is admissible and, further, that it is not appropriate to exclude the evidence in the exercise of my discretion.
0
7
1