Research paper
Transfer and persistence of non-self DNA on hands over time: Using empirical data to evaluate DNA evidence given activity level propositions

https://doi.org/10.1016/j.fsigen.2017.11.017Get rights and content

Highlights

  • Bayesian networks (BNs) are used to evaluate a pair of activity level propositions.

  • Probabilities on the persistence of non-self DNA on hands over time are obtained.

  • Delays in deposition decreased the detectability of the transferred contribution.

  • Empirical estimations of the likelihood of direct or indirect deposition are obtained.

  • Impacts of changes to various parameter values on the LR are observed.

Abstract

Questions relating to how DNA from an individual got to where it was recovered from and the activities associated with its pickup, retention and deposition are increasingly relevant to criminal investigations and judicial considerations. To address activity level propositions, investigators are typically required to assess the likelihood that DNA was transferred indirectly and not deposited through direct contact with an item or surface. By constructing a series of Bayesian networks, we demonstrate their use in assessing activity level propositions derived from a recent legal case involving the alleged secondary transfer of DNA to a surface following a handshaking event.

In the absence of data required to perform the assessment, a set of handshaking simulations were performed to obtain probabilities on the persistence of non-self DNA on the hands following a 40 min, 5 h or 8 h delay between the handshake and contact with the final surface (an axe handle). Variables such as time elapsed, and the activities performed and objects contacted between the handshake and contact with the axe handle, were also considered when assessing the DNA results.

DNA from a known contributor was transferred to the right hand of an opposing hand-shaker (as a depositor), and could be subsequently transferred to, and detected on, a surface contacted by the depositor 40 min to 5 h post-handshake. No non-self DNA from the known contributor was detected in deposits made 8 h post-handshake. DNA from the depositor was generally detected as the major or only contributor in the profiles generated. Contributions from the known contributor were minor, decreasing in presence and in the strength of support for inclusion as the time between the handshake and transfer event increased.

The construction of a series of Bayesian networks based on the case circumstances provided empirical estimations of the likelihood of direct or indirect deposition. The analyses and conclusions presented demonstrate both the complexity of activity level assessments concerning DNA evidence, and the power of Bayesian networks to visualise and explore the issues of interest for a given case.

Introduction

Over the last 15 years, there have been an increasing number of cases worldwide where DNA evidence has been questioned due to uncertainty about the activities that lead to its deposition, as in R v Hillier [1], R v Weller [2] and the case of the ‘Death of Meredith Kercher’ [3]. Several of these cases have been highlighted in the media, drawing significant attention to the methods applied to evaluate such evidence. While each case has different aspects that are unique to the scenario presented, the generic statements about the possibility of transfer do not adequately address the recurring questions being raised.

When considering propositions relating to activities of interest, termed activity level within the hierarchy of propositions [4], [5], the DNA findings are evaluated in light of how they support the competing propositions from prosecution and defence. Such evaluations often require the analyst to assess the likelihood that DNA was transferred indirectly (secondary transfer or further), compared to the possibility that it was transferred through direct contact with an item or surface. Knowledge of the factors affecting DNA transfer, persistence, prevalence and recovery (DNA-TPPR) within the context of a case are of importance to enable an accurate and case-specific assessment of the likelihood of biological examination results given particular activities. To inform the courts properly, analysts must draw upon DNA-TPPR data from relevant in-house and/or published studies, or, in the absence of objective data, expert elicitation, where the expert (with or without assistance from multiple experts) provides an opinion based on their knowledge and experience [6]. The authors advocate the use of objective data over expert elicitation, though a statement of limitations should accompany the use of data from either method.

While extensive research has/is being undertaken to extend our knowledge on DNA-TPPR in various scenarios, i.e. [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], the number of factors influencing transfer events, and high degree of variability within and between these factors, complicate activity level assessments [21]. However, this should not deter forensic practitioners from proceeding with such evaluations [22], [23].

Probabilistic methods have previously been demonstrated by others to be useful in addressing questions relating to activities [7], [15]. Furthermore, as a graphical representation of Bayesian formulations, Bayesian networks (BNs) simplify working with complex statistical formulae and have been used to address questions at source [24], [25], [26] and activity level [27], [28], [29]. The ease with which Bayesian networks can be used to incorporate multiple variables make them ideal in complex cases, enabling the investigation of a diverse range of scenarios and factors. The use of empirical data to inform parameters decreases subjectivity, while the ability to graphically model case-specific scenarios with clearly defined relationships and assumptions among variables, provides transparency. Furthermore, the ability to analyse the impact of assumptions and the sensitivity of the likelihood ratio (LR) to changes in variables, can inform the need and relevance of data or specific variables [30].

Here we use aspects of the “Fitzgerald case” from South Australia [31], [32] as a demonstration of how Bayesian networks can be used to evaluate DNA evidence when considering a pair of activity level propositions within the context of a case. This case was selected because of the numerous variables to be considered, the commonality of the factors and issues across cases, the various levels of uncertainty surrounding the activities involved and the fact that descriptions of activities deemed potentially relevant in the case were publically available. This case is primarily used as an example and it is not our intention to re-evaluate the evidence or the case specifically, but to simply demonstrate a conceptual framework for how certain aspects of cases like this might be evaluated in the future. As a result, there are many aspects of the case that are not mentioned or considered. Further, as the case was selected only as an exemplar of issues facing forensic biologists, the circumstances and variables have not been exactly replicated, but have been modelled within the research laboratory to provide indicative data in similar scenarios. As such, estimates of evidentiary weight should not be applied to the specific case utilised.

Here we provide a brief outline of the case circumstances. For further information see [31] and [32]. In the early hours of the morning, a large group (6+) of male offenders yielding weapons forced entry into a home that was being occupied by a number of individuals. Two of the occupants (D and K) were critically injured, with one (D) later dying in hospital. Sumner, who had been at the house visiting relatives prior to the attack, was charged with murder.

Along with other items, a didgeridoo found near the deceased was submitted for DNA testing. The didgeridoo, normally kept beside a washing machine in the laundry and used infrequently, was played by the deceased (D) just hours before the attack. There was no evidence as to how it came to be in the lounge or whether it was used in the attack.

A number of samples were collected (scrapings, swabs and tape-lifts) from bloodlike stains (visual and presumptive tests) on the didgeridoo and DNA profiles generated. A sample obtained from one end of the didgeridoo, which consisted of two blood-like spots, gave a mixed DNA profile that could be explained by a minimum of two contributors. Following a database search, the major component within the profile linked to Fitzgerald with an LR of 200 million; an unknown source contributed to the minor component. Fitzgerald’s contribution to this DNA profile was the only evidence linking him to the offence. Contributors to other profiles obtained from the didgeridoo consisted of the victims (D and K), frequently as major contributors, and unknown source/s as the minor component, although some profiles were too complex for interpretation or contained insufficient material to obtain a profile.

Fitzgerald claimed that he was not involved in the attack and had no association with the house or its occupants. He maintained that his DNA came to be on the didgeridoo through secondary transfer via Sumner, where earlier in the evening (∼8 h prior to the attack) he shook hands with Sumner at a boxing tournament, once as Sumner arrived and once as Sumner departed. Sumner, who arrived at the house ∼40 min after the boxing match, did not dispute the acquaintance with Fitzgerald or the handshakes at the boxing match. During his ∼5 h visit to the house, Sumner became involved in several altercations before leaving. Sumner’s presence during the attack several hours later was supported by witnesses and DNA evidence, though his DNA was not detected on the didgeridoo.

Despite Fitzgerald claiming he had no involvement in the offence and that his DNA had been secondarily transferred to the didgeridoo during the offence (8 h post-handshake), Fitzgerald was convicted of murder (along with Sumner) based on DNA evidence alone. An appeal some years later saw Fitzgerald acquitted, as the possibility of secondary transfer during Sumner’s first visit to the house could not be ruled out [32].

Both source and activity level propositions are of interest in this case as they were both raised at trial. In this paper, we aim to address questions at activity level. Hence, with matching DNA profiles we explicitly assume that a sample contains DNA of the person of interest (POI). Based on the timeline of events and points raised in the case, it was hypothesised that secondary transfer of Fitzgerald’s DNA via Sumner could have occurred on Sumner’s first visit to the house (between 40 min–5 h after the handshake with Fitzgerald), or on Sumner’s second visit to the house during the attack (8 h post-handshake).

We designed a series of Bayesian networks to deal with specific aspects of this case. During the identification of relevant parameters, it became evident that there was a lack of data relating to the persistence of transferred DNA on hands over time. As a result, we performed a number of handshaking experiments based on the time intervals presented in the case, and collected the data required to calculate the probabilities for use in the Bayesian networks. To provide a broader application to the data collected, an axe handle was used as a substitute for a didgeridoo. In addition, data collected from the timeframes observed by Szkuta et al. [17], namely, immediate deposition post-handshake and 15 min, were also incorporated to broaden the application of the Bayesian networks beyond this case. Furthermore, we explore the impact of changes to the various parameter values on the LR by performing sensitivity analyses. Based on these objectives, the structure of this paper follows the case investigation process, where, generally, one would first design a model, identify relevant parameters, and then collect the data needed to inform those.

Section snippets

Bayesian network construction

The Bayesian network in Fig. 1 (BN1) considers the core case circumstances outlined earlier and was built using Hugin Expert Software (Hugin Explorer v8.5). For more information on this software, we direct the reader to http://www.hugin.com/index.php/publications. A number of variables (nodes) with various dependencies (arrows) were considered during the construction of BN1. Starting at the ‘DNA profiling results’ node and working backwards, these results will inevitably be influenced by whose

Experimental design

As previously described by Szkuta et al. [17], different male (n = 11) and/or female (n = 25) volunteers who did not share the same living or working spaces were paired on eighteen separate occasions and a firm handshake performed for 10 s with their right hands (unwashed). In the current study, the handshake was repeated following a 10 s delay in order to simulate the general case circumstances. Following the two handshakes, individuals returned to their daily activities for an interval of 40 min (n = 

Quantification

A summary of the results obtained from the 108 samples collected from contacted axe handles and the right and left hands of the depositor 40 min, 5 h or 8 h after shaking hands with the known contributor are presented in Table 1 (refer to Supplementary Table 2 for detailed results). DNA quantities varied significantly between deposits on axe handles (t35 = 4.230, p < 0.001), between samples generated from the right hand of depositors (t34 = 5.430, p < 0.001) and between samples generated from the left

Conclusion

We present a series of Bayesian networks, each addressing different features specific to the R v Fitzgerald case, in order to evaluate a pair of activity level propositions. From its simplest form, considering a single secondary transfer event, to modelling two pathways of transfer and the absence of DNA from the vector, Bayesian networks can be used to evaluate DNA evidence when the activities surrounding its deposition are disputed. Performing experiments specific to the case scenario enabled

Acknowledgements

This work was supported by Deakin University and an Australian Government Research Training Program Scholarship.

References (48)

  • K. Steensma et al.

    An inter-laboratory comparison study on transfer, persistence and recovery of DNA from cable ties

    Forensic Sci. Int. Genet.

    (2017)
  • B. Szkuta et al.

    Transfer and persistence of DNA on the hands and the influence of activities performed

    Forensic Sci. Int. Genet.

    (2017)
  • M. van den Berge et al.

    Prevalence of human cell material: DNA and RNA profiling of public and private objects and after activity scenarios

    Forensic Sci. Int. Genet.

    (2016)
  • R.A.H. van Oorschot et al.

    Persistence of DNA deposited by the original user on objects after subsequent use by a second person

    Forensic Sci. Int. Genet.

    (2014)
  • T.J. Verdon et al.

    The influence of substrate on DNA transfer and extraction efficiency

    Forensic Sci. Int. Genet.

    (2013)
  • G. Meakin et al.

    DNA transfer: review and implications for casework

    Forensic Sci. Int. Genet.

    (2013)
  • J. de Zoete et al.

    Cell type determination and association with the DNA donor

    Forensic Sci. Int. Genet.

    (2016)
  • D. Taylor et al.

    Evaluating forensic biology results given source level propositions

    Forensic Sci. Int. Genet.

    (2016)
  • D. Taylor et al.

    Helping to distinguish primary from secondary transfer events for trace DNA

    Forensic Sci. Int. Genet.

    (2017)
  • D. Taylor et al.

    Using sensitivity analyses in Bayesian networks to highlight the impact of data paucity and direct future analyses: a contribution to the debate on measuring and reporting the precision of likelihood ratios

    Sci. Justice

    (2016)
  • I.W. Evett et al.

    More on the hierarchy of propositions: exploring the distinction between explanations and propositions

    Sci. Justice

    (2000)
  • T. Hicks et al.

    The importance of distinguishing information from evidence/observations when formulating propositions

    Sci. Justice

    (2015)
  • B. Kokshoorn et al.

    Activity level DNA evidence evaluation: on propositions addressing the actor or the activity

    Forensic Sci. Int.

    (2017)
  • J. de Zoete et al.

    Modelling crime linkage with Bayesian networks

    Sci. Justice

    (2015)
  • Cited by (49)

    • Improvements, factors, and influences on DNA recovery from firearms

      2023, Forensic Science International: Genetics
    View all citing articles on Scopus
    View full text