Research paper
Second GHEP-ISFG exercise for DVI: “DNA-led” victims’ identification in a simulated air crash

https://doi.org/10.1016/j.fsigen.2021.102527Get rights and content

Highlights

  • The GHEP-ISFG has designed a second “DNA-led” DVI exercise in a simulated air crash.

  • This exercise can be useful for testing DVI using the data and results provided.

  • The exercise is focused on direct matching and kinship analysis in a Bayesian framework.

  • This exercise sheds light on problems that a laboratory can face in DVI scenarios.

Abstract

The Spanish and Portuguese-Speaking Working Group of the International Society for Forensic Genetics (GHEP-ISFG) has organized a second collaborative exercise on a simulated case of Disaster Victim Identification (DVI), with the participation of eighteen laboratories. The exercise focused on the analysis of a simulated plane crash case of medium-size resulting in 66 victims with varying degrees of fragmentation of the bodies (with commingled remains). As an additional difficulty, this second exercise included 21 related victims belonging to 6 families among the 66 missings to be identified. A total number of 228 post-mortem samples were represented with aSTR and mtDNA profiles, with a proportion of partial aSTR profiles simulating charred remains. To perform the exercise, participants were provided with aSTR and mtDNA data of 51 reference pedigrees —some of which deficient—including 128 donors for identification purposes. The exercise consisted firstly in the comparison of the post-mortem genetic profiles in order to re-associate fragmented remains to the same individual and secondly in the identification of the re-associated remains by comparing aSTR and mtDNA profiles with reference pedigrees using pre-established thresholds to report a positive identification. Regarding the results of the post-mortem samples re-associations, only a small number of discrepancies among participants were detected, all of which were from just a few labs. However, in the identification process by kinship analysis with family references, there were more discrepancies in comparison to the correct results. The identification results of single victims yielded fewer problems than the identification of multiple related victims within the same family groups. Several reasons for the discrepant results were detected: a) the identity/non-identity hypotheses were sometimes wrongly expressed in the likelihood ratio calculations, b) some laboratories failed to use all family references to report the DNA match, c) In families with several related victims, some laboratories firstly identified some victims and then unnecessarily used their genetic information to identify the remaining victims within the family, d) some laboratories did not correctly use “prior odds” values for the Bayesian treatment of the episode for both post-mortem/post-mortem re-associations as well as the ante-mortem/post-mortem comparisons to evaluate the probability of identity. For some of the above reasons, certain laboratories failed to identify some victims. This simulated “DNA-led” identification exercise may help forensic genetic laboratories to gain experience and expertize for DVI or MPI in using genetic data and comparing their own results with the ones in this collaborative exercise.

Introduction

The identification of missing persons in large-scale events such as disasters (DVI) or, for example, mass graves from past armed conflicts, is a challenge for forensic services because of the complexity that the context may present, which can profoundly influence the difficulty for the correct identification of the victims [1]. Several best practice and procedure recommendations have been published regarding DVI [1], [2] and missing persons identification (MPI) [3], [4] investigations. The ISFG has also issued recommendations for forensic genetic laboratories in order to help them to deal with the identification process in the context of large number of victims [5].

Several factors can influence the complexity of the identification process in DVI or MPI scenarios [1]. One of them is the number of missing persons —a large number of victims has deep influence on the Bayesian framework of the context and, therefore, on the capability to correctly identify the victims. In addition, the degree of the disarticulation may influence the number of DNA tests required for DNA-based re-association of post-mortem samples. DNA degradation can influence the quality of the retrieved genetic information from the human remains. Autosomal STR markers (aSTRs) are powerful systems to build genetic profile databases for DVI or MPI due to their high discrimination and individualization power [1], [2], [3], [4], however partial genetic profiles with extensive allele or locus dropout due to DNA degradation can considerably reduce the power of discrimination of these DNA profiles, which may be even more problematic if multiple close relatives are not available to profile reference samples.

The existence of several related victims among the missing is another factor that can considerably complicate the DNA identification process [6], [7]. Regarding reference samples for genetic comparisons, the best DNA sources are victim’s ante-mortem biological specimens due to the high power of identification through direct genetic comparisons [6]. Nevertheless, sometimes there is some uncertainty about the real origin of profiles recovered from personal belongings of missing persons. In addition, it may not be possible to obtain ante-mortem biological material from victims as, for example, in cases of mass graves from human rights violations that are investigated long after the event. In such cases, samples from victims’ relatives become the appropriate source of genetic in order to carry out the identification process. Hence, the quality of family pedigrees may deeply influence the success of the identifications: deficient pedigrees made up of few first-degree or only second/ third-degree relatives may diminish to a great extent the power of identification [7] and may even prevent victim identification by producing weak evidence that is difficult to distinguish from adventitious matches to unrelated individuals [6], [7]. In the latter case, the analysis and comparison of lineage markers such as Y-STRs or mtDNA may prove useful to guide the identification of victims having only distant relatives. However, the informativeness of lineage markers is limited when there are related victims belonging to the same lineage. In kinship analyses, inconsistencies in reported family relationships (for example, incidental findings of non-paternity) may hinder the identification of the victim, imposing the need to re-analyze the family pedigree under different hypotheses of relatedness.

A proper Bayesian approach to large-scale identifications is based on likelihood ratios coming from DNA comparisons involving paired hypotheses (typically, but not always, the hypothesis of related versus unrelated), multiplied by the prior odds for an identification (typically the inverse of the number of missing persons in an event). A DVI episode may be classified as a “closed” event, for example an air crash in which the number and identity of the missing persons are known, making it a simple matter to define the prior odds as the inverse of the known number of missing persons. In closed events, prior odds can be refined further by considering other contextual or non-DNA evidence, such as age, sex, location, etc. Within other contexts of MPI, for example post-conflict mass graves or enforced disappearances, disappearances may accumulate over time and in different places, and there may be less ante-mortem non-genetic information available, and the event may be “open”, without a well-defined number of missing persons. In such open events, defining appropriate prior odds can be more complex, requiring some form of reasoned and operational prior probability to be established considering the context. Furthermore, forensic teams can establish minimum statistical thresholds to consider an identification as reliable, depending on the context of the event [5].

The GHEP-ISFG has previously carried out a simulated MPI collaborative exercise requiring participants to perform bone re-associations and identification of missing persons in a secondary common grave with commingled remains; 11 laboratories participated and there were several lessons learned [8]. In keeping with its interest in collaborative exercises for human identification under DVI or MPI contexts, the GHEP-ISFG organized and documented a second simulated DNA-led exercise within the context of a medium-scale disaster, the results of which are reported here. This simulated scenario was more complex than in the first exercise, including complexity factors such as: fragmented remains requiring re-association through direct post-mortem/post mortem (pm-pm) comparison, partial aSTR profiles due to degraded DNA, ante-mortem/post-mortem (am-pm) comparisons using family references with diverse pedigrees, related victims belonging to the same families, family inconsistencies attributable to mutations, DNA match values below the established statistical threshold, and the requirement to consider mtDNA information to solve matches below threshold.

The goal of this second simulated "DNA-led" collaborative GHEP-ISFG DVI exercise is to continue learning and gaining experience in the comparison of genetic profiles in DVI / MPI contexts. In addition, this exercise may contribute to laboratories interested in preparing themselves in DVI or MPI, permitting a comparison of their results with the consensus obtained by the participating laboratories and highlighting the possible errors that can be made.

Section snippets

Autosomal STR (aSTR) profiles

Genetic profiles for 18 aSTR markers were simulated according to allele frequencies used by GHEP-ISFG in its annual inter-laboratory comparison exercises [9], using the Familias software [10]. To this end, a large family pedigree of 14 individuals including three generations was designed (Fig. 1A and B) and 1000 simulations were performed to obtain the aSTR genotypes for each individual within the pedigree. Once the aSTR profiles in each complete pedigree chart were simulated, a custom Python

Scenario description for participants

The present exercise simulates an air crash with victims whose remains are fragmented. The flight list describes the presence of 66 victims including passengers and crew and defines the disaster as a “closed” case. As post-mortem samples (pm) presented various degrees of preservation (some of them were burnt), many of them yielded partial profiles. The agency in charge of coordinating forensic tasks and contacting the victims' relatives gathered biological samples of 128 family references.

Aim 1 - pm-pm comparisons: re-association of profiles

A direct comparison of genetic profiles from pm samples was requested in order to re-associate pm samples to an individual.

A likelihood ratio (LR) value equal to or greater than 1.0E + 07 (LR  10,000,000) was defined as a reliable re-association threshold among pm profiles but laboratories were instructed to not report LR values for pm-pm re-associations but just to group them.

Once the pm samples/profiles were re-associated, participants were requested to group pm samples indicating the code of

Pm-pm comparisons for re-association of “fragmented remains”

The first objective of this exercise was to compare the 228 pm profiles for the re-association of fragmented remains, provided that the LR threshold to report a pm-pm DNA match was LR ≥ 1.0E + 07. No pm-pm DNA matches yielding a LR value lower than 1.0E + 07 were included in the exercise, therefore all the pm profiles, in theory, could be re-associated into 65 single profiles.

Regarding the pm-pm re-associations, a divergent result was considered when at least one of the re-associated fragments

Discussion and conclusions

This simulated DVI exercise considered different complexities that forensic genetic laboratories may face when massively comparing genetic profiles in order to identify victims in a disaster of these characteristics. The exercise was designed as a "DNA-led" DVI project allowing the identification of all the victims with the exception of the group pm1842/pm1959. These pm profiles did not match any reference family, while two families, F82 and F88, did not match any remains’ profile. In addition,

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Acknowledgments

We thank the 18 laboratories that participated in this exercise.

References (28)

  • D. Kling et al.

    Familias 3—extensions and new functionality

    Forensic Sci. Int. Genet.

    (2014)
  • C.H. Brenner et al.

    Issues and strategies in the identification of World Trade Center victims

    Theor. Pop. Biol.

    (2003)
  • C.H. Brenner

    Some mathematical problems in the DNA identification of victims in the 2004 tsunami and similar mass fatalities

    Forensic Sci. Int.

    (2006)
  • A. Sozer, M. Baird, M. Beckwith, B. Harmon, D. Lee, G. Riley, S. Schmitt, Guidelines for Mass Fatality DNA...
  • Cited by (8)

    • Effects of DNA degradation and genotype imputation on high-density SNP microarray in pairwise kinship analysis

      2023, Legal Medicine
      Citation Excerpt :

      Kinship analysis is, for example, utilized in paternity testing, missing person identification, or disaster victim identification (DVI). In DVI, post-mortem samples and ante-mortem samples are collected from unidentified remains (e.g., bones, teeth, or nails) and family references, respectively [1,2]. Relationships between the samples are determined by calculating likelihood ratios (LR) or posterior probability.

    • Recent advances in forensic biology and forensic DNA typing: INTERPOL review 2019–2022

      2023, Forensic Science International: Synergy
      Citation Excerpt :

      A simulated airplane crash enabled six forensic laboratories in Switzerland to gain valuable DVI experience with kinship cases of varying complexity [600]. The ISFG Spanish-Portuguese Speaking Working Group likewise conducted a DVI collaborative exercise with a simulated airplane crash to explore fragment re-associations, victim identification through kinship analysis, coping with related victims, handling mutations or insufficient number of family references, working in a Bayesian framework, and the correct use of DVI software [601]. Other groups have explored the capability of a particular software tool [602] or implemented rapid DNA analysis to accelerate victim identification [603].

    View all citing articles on Scopus
    View full text