WK5: DNA EVIDENCE

Forensicgeneticgenealogyusingmicroarraysfortheidentificationofhumanremains_TheneedforgoodqualitysamplesApilotstudy.pdf

Home >Law homework help >Criminal homework help >WK5: DNA EVIDENCE

Contents lists available at ScienceDirect

Forensic Science International

journal homepage: www.elsevier.com/locate/forsciint

Forensic genetic genealogy using microarrays for the identification of human remains: The need for good quality samples – A pilot study

A. Davawalaa, A. Stocka, M. Spidena, R. Danielc, J. McBaind, D. Hartmana,b,⁎

a Victorian Institute of Forensic Medicine, Victoria, Australia b Department of Forensic Medicine, Monash University, Victoria, Australia c Office of the Chief Forensic Scientist, Victoria Police Forensic Services Department, Victoria, Australia d totheletterDNA, Brisbane, Queensland, Australia

a r t i c l e i n f o

Article history: Received 2 December 2021 Received in revised form 31 January 2022 Accepted 23 February 2022 Available online 25 February 2022

Keywords: DNA analysis Forensic Genetic Genealogy Missing persons

a b s t r a c t

The successful application of forensic genetic genealogy (FGG) to identify Jane and John Doe cases in the United States has raised the prospect of using the technique in Australia to assist in the reconciliation of unidentified human remains (UHRs) with long term missing persons. A study was conducted to explore the feasibility of FGG using whole genome array (WGA) data from both pristine control samples as well as compromised casework samples, with the view to explore how DNA quantity and quality impacted on the ability to generate search results when compared to a genetic genealogy database, such as GEDmatch. From this study, several insights were gained as to the impact DNA quantity and degradation had on the per- centage of SNPs genotyped and heterozygote/homozygote ratio – which are critical for successful matching outcomes. It was noted in this study (using a control sample) that successful matching occurred when genotyping errors were 5% or less. Two UHR cases were matched to kits on GEDmatch PRO, which provided investigative leads for identification purposes. The effectiveness of the FGG approach to match casework samples (as well as volunteer samples used in the study) is indicative of the usage of ‘direct-to-consumer’ (DTC) genetic testing by Australians. Given the (often) limited availability of casework samples, findings from this study will assist Australian agencies considering the use of FGG, to determine if WGA is a suitable method for their application.

1. Introduction

The identification of a deceased person is an integral part of the medico-legal death investigation process in Victoria, Australia, with unidentified human remains (UHR) cases reported to the Coroner for investigation [1]. For these cases, the use of DNA matching is reliant on 1) the ability to recover a DNA profile from the deceased; and 2) the availability of a comparison sample (either from the missing person or a relative). From a DNA point of view, if the DNA profile from a UHR case is not matched in a local, state or national missing persons DNA database, an identification cannot be achieved and must wait for the appropriate Ante-Mortem (AM) data to be made available. In these instances, forensic genetic genealogy (FGG) – an emerging field for forensic investigation – has the potential to pro- vide an alternate avenue for identification [2–6].

FGG, also known as Investigative Genetic Genealogy (IGG), combines DNA testing – using Whole Genome Array (WGA) or Whole Genome Sequencing (WGS) data – with traditional genealo- gical methods to infer (familial) relationships between individuals [3]. Unlike conventional Short Tandem Repeat (STR) analysis, which is only useful for matching to close relatives, FGG will enable mat- ches extending to more distant relatives (such as 1st, 2nd, or 3rd cousins) due to the large number of DNA sites that are available for comparison. This is achieved by determining the DNA segments shared by individuals – shared centimorgan (cM) – with the more segments you share with someone, the more closely related you are. The familial connections can then be used to provide leads for the purpose of identification.

Our current awareness regarding the use of FGG for identification purposes has come from law enforcement (LE) and other agencies and/or projects in the United States (U.S.). From a human identifi- cation point of view, most notable is the DNA Doe project, a non- profit organisation, which aims to identify Jane and John Does in the U.S. using FGG through donations and volunteer work of genetic

]]]] ]]]]]]

⁎ Corresponding author at: Victorian Institute of Forensic Medicine, Victoria, Australia.

E-mail address: [email protected] (D. Hartman).

Forensic Science International 334 (2022) 111242

genealogists [7]; with early success stories including the an- nouncement in April 2018 that a Jane Doe discovered in Ohio U.S., who became known as the Buckskin Girl, had been identified using this approach [8]. However, it was the use of FGG to identify the Golden State Killer [6] that really thrust FGG into the public domain. Such is the interest in the use of FGG, that companies (such as Parabon NanoLabs [9], Bode [10], and Othram [11]) are offering FGG as a service to assist with cold case and missing persons investiga- tions. Nevertheless, further work is still required not only to review the scientific and technical aspects of FGG, such as in de Vries et. al. [12], but also consider the ethical, privacy and legal factors sur- rounding FGG, given the use of public database held by direct-to- consumer (DTC) companies, and the creation of large data sets [13–15]. All these factors will impact on how FGG would function once operationalised, with forensic agencies needing to consider all of them in the context of their legislative frameworks.

FGG is in its infancy in Australia, with cases being considered for FGG particularly when all other avenues for identification have been exhausted [16]. Given that there is usually a finite amount of DNA available for analysis, it is imperative that the likelihood of success based on this approach is understood, as well as the limitations that compromised casework samples present. To this end, a pilot study was undertaken to evaluate FGG using both good quality control samples as well as casework samples from UHR cases, with the aim of providing preliminary information regarding DNA quality and ability to generate a match list when searched against a genealogy database, such as GEDmatch/GEDmatch PRO. GEDmatch/GEDmatch PRO enables the comparison of DNA data from control and casework samples to DNA data from users that have utilised one of several DTC companies and have uploaded their data to GEDmatch. Matches are categorised based on the total shared DNA segments (shared cM values) between the questioned sample to the nearest matches, with the greater amount of shared DNA indicative of a closer familial relationship. DNA samples recovered from the control and casework samples were subjected to WGA using the OmniExpress microarray (previously routinely used by DTC companies). Data was uploaded to GEDmatch (control samples) or GEDmatch PRO (casework samples), with one casework sample further investigated using Family- TreeDNA. The impact of DNA quantity and quality on the ability to generate SNP data suitable for upload and comparison on genetic genealogy databases were evaluated.

2. Methods

2.1. Sample Selection

Samples were collected (and analysed) with approval from the Victorian Institute of Forensic Medicine Ethics Committee, EC 11–2019. As part of the approval process considerations were given as to the use of casework samples – for which all current avenues for

identification had been exhausted – including data storage and analysis, as well as privacy and ethical matters.

2.1.1. Ideal (Positive Control) Sample A control sample (buccal swab or toenail clipping) was obtained

from a living donor who has known kits on GEDmatch. This sample, herein referred to as ‘ideal sample’, was expected to yield good quality DNA (based on the degradation index (DI)) in enough quantity for optimal processing on the microarray. Consent was provided by the donor for their DNA to be genotyped and uploaded to GEDmatch.

2.1.2. Casework Samples Eight unidentified human remains cases were selected for FGG

(Table 1). The post-mortem interval (PMI) could only be estimated for two cases, as most (75%) were skeletal remains of unknown antiquity. All cases had conventional nuclear DNA profiling (using Identifiler™ Plus or GlobalFiler™ on a 3500 genetic analyser, Ther- moFisher Scientific) and mitochondrial DNA profiling (using Sanger sequencing on a 3500 genetic analyser) with either complete or partial DNA profiles available for identification purposes (Table 1). Comparisons to the Victorian Missing Persons DNA Database (VMPDD) – which at present holds more than 350 reference samples for missing persons – failed to produce a familial match in these cases. Where possible, biogeographical ancestry (BGA) predictions (previously derived using the Precision ID Ancestry Panel, Thermo- Fisher Scientific) were used to assist with case selection (Table 1), as subjects of European ancestry have a higher probability of success due to their over-representation in genetic genealogy databases [4]. Sample types were mostly bone samples, with one sample being a bloodstain sample (Table 1).

2.1.3. Control Samples Four control samples were collected from laboratory staff in-

volved in the processing of the casework samples for elimination purposes. These samples were either toenail clipping (x3) or skin (x1) samples. Toenail clippings were collected as a sample type, as this is a sample often analysed by the laboratory for casework, while the skin sample provided an additional sample type to analyse. Consent was provided by the donors for their DNA to be genotyped and uploaded to GEDmatch.

2.2. DNA Extractions

DNA was extracted from the ideal sample (buccal swab or toenail clipping) and other control samples (toenail clipping or skin) sam- ples using a QIAamp® DNA investigator kit (Qiagen) on a QIAcube platform following the manufacturer’s protocols, with an elution volume of 100 μL.

Table 1 Case selection – unidentified human remains.

Case No. Age of case (yrs)a

Body condition PMI Sample type for DNA extraction

nDNA profile available (No. loci)b

mtDNA profile available (HVI and/or HVII)

BGA prediction availability

1 17 Intact Days Bloodstain 15/15 HVI & HVII Yes, European 2 14 Mandible with teeth Unknown Bone 15/15 HVI & HVII Yes, European 3 31 Skeletal (complete) Unknown Bone 13/15 HVI & HVII Yes, European 4 15 Skeletal (skull and other

bones) Unknown Bone 13/15 HVI & HVII Yes, European

5 26 Skeletal (skull and other bones)

Unknown Bone 15/15 HVI & HVII Yes, European

6 4 Skeletal (complete) Unknown Bone 21/21 HVI & HVII Yes, European 7 11 Skeletal (humerus) Unknown Bone 12/15 HVI & HVII Yes, European 8 1 Decomposed (complete) Days-month Bone 21/21 HVI & HVII No

a time since case reported – does not reflect the age of the deceased or the time since death. b Number of loci (not including sex determination markers) for Identifiler™ Plus or GlobalFiler™ out of 15 or 21 respectively.

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

DNA was extracted from the bloodstain casework sample (Case 1) using a QIAamp® DNA investigator kit (Qiagen) completed manually, rather than on a robotic platform, following the manufacturer’s protocols, with a total elution volume of 100 μL (4 ×25 μL).

For the bone casework samples, for the first extraction run, DNA was extracted from casework samples (cases 2–7) using a phenol/ chloroform DNA extraction method [17] using an input amount of 1–4 g of bone powder. For the second extraction run, four of the bone samples (cases 3, 4, 5 and 6), as well as Case 8, were extracted using higher quantities of bone powder (15–30 g of bone powder) and including a demineralisation step (demineralisation buffer - 0.5 M EDTA, pH 8.0, 1% (w/v) N-Lauroylsarcosine) prior to the phenol/chloroform extraction. The demineralisation buffer was used at a ratio of 15 mL for every 1 g of bone and the sample was placed in a rotating incubator at 56 °C for 12–24 hrs.

All sample extractions included extraction blanks to detect any laboratory contamination.

2.3. DNA Quantification and DNA Genotyping

Nuclear DNA concentration was determined using Quantifiler™ HP (Life Technologies) using a 7500 real-time PCR instrument (ThermoFisher Scientific) under conditions specified by the manu- facturer [18].

DNA extracts were genotyped using an Illumina OmniExpress- 24 [19] at the Australian Genome Research Facility (AGRF) under conditions specified by the manufacturer. Samples were run on two chips with varying input amounts (Table 2). Two analyses were conducted, the first consisted of running samples VIFM-01 - VIFM- 24 on an OmniEpress-24 microarray (chip 1), and the second con- sisting of samples VIFM-25 - VIFM-48 on a second OmniExpress-24 microarray (chip 2). Cases 1-7 were run in duplicate on chip 1, while cases 3–6 as well as Case 8 were run in duplicate on chip 2, resulting in cases 3–6 not only having intra-chips duplicates, but also inter-chip duplicates (Table 2). VIFM-01 was a positive control

Table 2 SNP analysis – Samples analysed on the Illumina OmniExpress-24 microarray. Samples VIFM-01 - VIFM-24 were run on the chip 1 and samples VIFM-25 - VIFM-48 were run on chip 2. Gender and call rate obtained from the SNP analysis is noted.

Sample ID Sample Sample Type Concentration (ng//μL) Deg Index Input (ng) Gender Call Rate

VIFM-01 Ideal Sample Buccal swab 38.4096 0.83 153.6384 √ 0.99 VIFM-02 Ideal Sample Buccal swab 38.4096 0.83 80.0000 √ 0.99 VIFM-03 Ideal Sample Buccal swab 38.4096 0.83 40.0000 √ 0.99 VIFM-04 Ideal Sample Buccal swab 38.4096 0.83 4.0000 √ 0.98 VIFM-05 Ideal Sample Buccal swab 38.4096 0.83 0.4000 Unknown 0.94 VIFM-06 Ideal Sample Buccal swab 38.4096 0.83 0.0400 Unknown 0.66 VIFM-07 Ideal Sample Buccal swab 38.4096 0.83 0.0040 Unknown 0.26 VIFM-08 Ideal Sample Buccal swab 38.4096 0.83 0.0004 Unknown 0.74 VIFM-09 Case 1 Bloodstain 3.1455 0.62 12.5820 √ 0.99 VIFM-10 Case 1 Bloodstain 3.1455 0.62 12.5820 √ 0.98 VIFM-11 Case 2 Tooth/mandible 0.0091 2.09 0.0364 Unknown 0.71 VIFM-12 Case 2 Tooth/mandible 0.0091 2.09 0.0364 Unknown 0.20 VIFM-13 Case 3 Bone 0.0617 1.15 0.2468 Unknown 0.52 VIFM-14 Case 3 Bone 0.0617 1.15 0.2468 NR NR VIFM-15 Case 4 Bone 0.0260 2.51 0.1040 Unknown 0.73 VIFM-16 Case 4 Bone 0.0260 2.51 0.1040 Unknown 0.66 VIFM-17 Case 5 Bone 0.1003 1.48 0.4012 Unknown 0.48 VIFM-18 Case 5 Bone 0.1003 1.48 0.4012 NR NR VIFM-19 Case 6 Bone 0.0108 1.61 0.0432 NR NR VIFM-20 Case 6 Bone 0.0108 1.61 0.0432 Unknown 0.27 VIFM-21 Case 7 Bone 0.0347 1.74 0.1388 NR NR VIFM-22 Case 7 Bone 0.0347 1.74 0.1388 NR NR VIFM-23 Mixture 5.8570 N/A 23.4280 Unknown 0.75 VIFM-24 Control -ve Unknown 0.22 VIFM-25 Ideal Sample Buccal swab 5.7000 1.01 22.8000 √ 0.98 VIFM-26 Ideal Sample Buccal swab 5.7000 1.01 0.9000 √ 0.97 VIFM-27 Ideal Sample Buccal swab 5.7000 1.01 0.8000 √ 0.98 VIFM-28 Ideal Sample Buccal swab 5.7000 1.01 0.7000 √ 0.97 VIFM-29 Ideal Sample Buccal swab 5.7000 1.01 0.6000 Unknown 0.96 VIFM-30 Ideal Sample Buccal swab 5.7000 1.01 0.5000 Unknown 0.95 VIFM-31 Ideal Sample Buccal swab 5.7000 1.01 0.1000 Unknown 0.88 VIFM-32 Ideal Sample Buccal swab 5.7000 1.01 0.0400 Unknown 0.79 VIFM-33 Ideal Sample Buccal swab 5.7000 1.01 0.0200 Unknown 0.64 VIFM-34 Case 5 Bone 0.1533 VD 0.6132 Unknown 0.88 VIFM-35 Case 5 Bone 0.1533 VD 0.6132 Unknown 0.89 VIFM-36 Case 3 Bone 0.1239 1.40 0.4955 Unknown 0.91 VIFM-37 Case 3 Bone 0.1239 1.40 0.4955 Unknown 0.93 VIFM-38 Case 4 Bone 0.5390 0.97 2.1560 Unknown 0.70 VIFM-39 Case 4 Bone 0.5390 0.97 2.1560 Unknown 0.77 VIFM-40 Case 6 Bone 0.2596 3.56 1.0384 Unknown 0.91 VIFM-41 Case 6 Bone 0.2596 3.56 1.0384 Unknown 0.90 VIFM-42 Case 8 Bone 3.4754 1.36 13.9015 Unknown 0.85 VIFM-43 Case 8 Bone 3.4754 1.36 13.9015 Unknown 0.87 VIFM-44 Ideal Sample Toenail 5.7277 2.64 22.9108 √ 0.97 VIFM-45 C-sample 1 Toenail 8.9482 2.30 35.7929 √ 0.98 VIFM-46 C-sample 2 Toenail 1.6843 4.96 6.7373 Unknown 0.64 VIFM-47 C-sample 3 Toenail 10.4638 1.88 41.8553 √ 0.98 VIFM-48 C-sample 4 Skin 16.2860 1.55 65.1439 Unknown 0.97

NR: no result obtained √: gender call as expected for the donor VD: very degraded

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

sample while VIFM-24 was a negative control (no DNA added) sample.

Data analysis for the Illumina OmniExpress-24 runs was con- ducted by AGRF using Illumina’s GenomeStudio 2.0.3 with Genotyping module 2.0.3, using the default Illumina settings and Illumina InfiniumOmniExpress-24v1–3_A1 manifest and Infinium OmniExpress-24v1–3-A1_ClusterFIle cluster files.

2.4. GEDmatch and GEDmatch PRO Uploads

Data files were prepared for upload to GEDmatch or GEDmatch PRO using Illumina’s GenomeStudio 2.0. An account was created under GEDmatch and GEDmatch PRO, respectively, with ideal and control samples uploaded to GEDmatch, and although casework samples were originally uploaded to GEDmatch these were moved to GEDmatch PRO when the terms and conditions for GEDmatch were updated in January 2021. While the VIFM is not a LE agency, GEDmatch now requires all kits uploaded for the purpose of iden- tifying human remains to be processed using GEDmatch PRO. Each data file uploaded to GEDmatch was assigned a unique kit identifier for subsequent searching, with all kits having a status of ‘research’ to ensure that the kit’s DNA data would not be included in match re- sults of other users [20]. Uploads to GEDmatch PRO were similarly assigned a unique kit identifier for subsequent searching.

For the casework samples that returned a poor match result – defined as either returning no matches or matching known ‘junk’ (artificial) kits – on GEDmatch PRO, additional data files were pre- pared (as described above) for upload following data clean up (see section below).

2.5. GEDmatch and GEDmatch PRO Searching

For GEDmatch, the subscription fee was paid to enable access to a number of on-line search tools. On GEDmatch, kits derived from the ideal and control samples were queried against the entire data set on the database (‘one-to-many’ comparison). More detailed one-to-one comparisons were done using the ‘one-to-one’ comparison tool; where the kit with the highest DNA input amount (150 ng) of the ideal sample was designated as having the correct SNP calls, to which all other ideal samples (with DNA input of 80 ng or less) would be compared to evaluate match outcomes. Population admixture pro- portions were estimated using the Eurogenes K13 model [21] offered as part of the GEDmatch tool kit. On GEDmatch PRO, kits derived from casework samples were queried against all kits that have opted-in for law-enforcement searches (‘one-to-many’ segment based comparison) with population admixture proportions estimated using the Eurogenes K13 model [21] offered as part of the GEDmatch PRO tool kit.

For all matches observed on GEDmatch and GEDmatch PRO, the shared cM values were used to determine likely relationship to the kit/sample in question using the Shared cM project v4 tool in DNA Painter (https://dnapainter.com/tools/sharedcmv4). This was the preferred method for determining relationships based on current best practice (as advised by genetic genealogist) rather than using the ‘Gen’ value provided by GEDmatch or GEDmatch PRO. Matches from the ‘one-to-many’ comparisons in GEDmatch Pro were cate- gorised based on the total shared DNA segments (shared cM values) between the sample and the nearest matches, with the greater amount of shared DNA indicative of a closer relationship.

At the completion of the study, 13th October 2021, the kits for the ideal and control samples were removed from GEDmatch.

2.6. FamilyTreeDNA Upload

The data file for Case 1 (VIFM-09) was submitted to a DTC da- tabase – FamilyTreeDNA – for comparison using the Gene By Gene / FamilyTreeDNA investigative Genetic Genealogy (IGG) Services [22].

2.7. Bioinformatics

2.7.1. Data Processing Data analysed using Illumina’s GenomeStudio 2.0.3 as described

in Section 2.3 Section 2.4 was further processed using Python pro- gramming language-based scripts for both data exploration and generation of composite SNP profiles.

2.7.2. Heterozygote/Homozygote Ratio The Heterozygote SNPs/Homozygote SNPs (Het/Hom) ratios were

calculated as a ratio of the number of SNPs that were heterozygous and those that were homozygous, using all SNPs that yielded gen- otyping data for all the samples.

2.7.3. SNP Concordance, Drop-out and Drop-in To evaluate concordance of SNP calls, a comparison was under-

taken for all dilutions of the ideal sample that yielded genotyping data. VIFM-01, with the highest DNA input amount (150 ng) was designated as the ideal sample, to which all other samples (with DNA input of 80 ng or less) would be compared to evaluate SNP call concordance, drop-out, and drop-in. Any SNPs which had no data available were removed prior to the concordance or drop-out eva- luations.

2.7.4. Genotyping Errors The impact of DNA input on types of genotyping errors was

evaluated. The types of genotyping errors considered were: i) false genotype – homozygote (both reference and comparison genotypes homozygote with different allele calls); ii) allele drop-out (false homozygote); and iii) allele drop-in (false heterozygote). The geno- typing error was calculated as a percentage of total SNP calls for the sample.

2.7.5. Composite Profiles Composite profiles were generated for the casework samples that

failed to yield a search outcome with the aim of cleaning and en- riching data for upload and searching. Described below are the manipulations of the SNP data undertaken. SNP data for cases 3 (VIFM-13, VIFM-14, VIFM-36 and VIFM-37), 4 (VIFM-15, VIFM-16, VIFM-38 and VIFM-39), 5 (VIFM-17, VIFM-34 and VIFM-35) and 6 (VIFM-20, VIFM-40 and VIFM-41) were interrogated and three new data files were created which had: (i) the removal non-concordant SNPs, as well as SNPs which had no data available, (ii) as (i) with the additional removal of all homozygote SNPs, and (iii) as (i) with the additional removal of all homozygote SNPs not seen in at least two of the replicates, created for each case.

3. Results

3.1. DNA Extracts & Genotyping

For the DNA extracts described below, SNP genotyping on mi- croarrays was conducted by an external service provider (AGRF). Reports detailing the performance of the two microarray runs were provided, with both runs passing all the required controls including the non-specific binding controls that target bacterial sequences (data not shown).

3.1.1. Ideal (Positive Control) Sample Two DNA extracts were prepared from buccal swabs with con-

centration (ng/μL)/ DI values of 38.4/0.83 (extract 1) and 5.70/1.01 (extract 2) respectively. An additional DNA extract was prepared from a toenail clipping with a concentration of 5.73 ng/μL and DI value of 2.64. Results from the SNP analysis are shown in Table 2. Extract 1 was used to prepare samples VIFM-01 to VIFM-08 for the first microarray run, with DNA input ranging in amounts from

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

approximately 150 ng to 0.4 pg. Extract 2 was used to prepare samples VIFM-25-VIFM-33 for the second microarray run ranging in input amounts from approximately 22 ng to 0.02 ng. The DNA ex- tract derived from the toenail clipping was analysed on the second microarray run with an input amount of 23 ng. Samples with an input amount ranging from 150 ng to 0.7 ng all had call rates > 0.97 and were able to provide the expected gender (data not shown). Samples with 0.6 ng to 0.4 ng input amounts had call rates > 0.93 and were not able to provide the expected gender. While samples with 0.1 ng to 0.04 ng input had call rates between 0.66 and 0.88, and samples with 0.02 ng or less input having call rates < 0.74.

3.1.2. Control Samples Three of the four control samples (C-samples) were DNA extracts

recovered from toenail samples with concentration (ng/μL)/DI values of: 8.9/2.30 (C-sample 1), 1.7/4.96 (C-sample 2) and 10.5/1.88 (C- sample 3); whereas the fourth DNA extract was recovered from a skin sample with a concentration of 16.3 ng/μL and DI value of 1.55. Results from the SNP analysis are shown in Table 2. Three of the four samples had call rates > 0.97 (C-samples 1, 3 and 4), with one sample (C-sample 2) having a call rate of 0.64. Two of the four samples (C-samples 1 and 3) with call rates > 0.976 were able to provide the expected gender for those samples, with the other two samples (C-samples 2 and 4) yielded an ‘unknown’ gender result.

3.1.3. Casework Samples The DNA extracts for the casework samples ranged in con-

centration from 0.009 ng/μL to 3.5 ng/μL with DI values ranging from 0.6 to 3.6, with results from the SNP analysis shown in Table 2.

In the first SNP analysis with Case 1-7, only one case (Case 1) had a call rate > 0.9, with some returning poorer call rates of < 0.73 in either one (cases 3, 5 and 6) or both replicates (cases 2 and 4), and some failing to yield a result in either one or both replicates, such as case 7. Only one case (Case 1) provided gender determination, all other cases yielded an ‘unknown’ gender result.

In the second SNP analysis, where samples of additional extrac- tions for cases 3–5 and an extra case not previously analysed (Case 8) were run, two cases (cases 3 and 6) had a call rate > 0.9; with the other cases yielding call rates between 0.7 and 0.8. None of the cases were able to provide gender determination and were ‘unknown’. For casework samples, the expected gender was known from STR pro- filing (data not shown).

3.1.4. Sample Quality For the samples that yielded SNP data, an analysis of how the

samples performed – the percentage of SNPs genotyped and Het/ Hom ratios – was undertaken. Of the non-degraded samples (ideal sample dilution series), DNA input amount was seen to correlate with percentage of SNPs genotyped and Het/Hom ratio (Fig. 1). When evaluating the combined impact of sample input amount and degradation on Het/Hom ratios, for samples with an input amount > 1 ng of DNA in the SNP assay degradation did not appear to have an adverse impact on the Het/Hom ratio; while for samples with an input amount of < 1 ng, degradation did appear to have an adverse impact on the Het/Hom ratio (Fig. 2). Evaluation of the impact of DNA input amount and degradation on the percentage of SNPs genotyped showed two clusters with a small overlap; one of samples that yielded > 85% SNPs genotyped, and one of samples that yielded < 85% SNPs genotyped (Fig. 2). Furthermore, when evalu- ating DNA input amounts to GEDmatch outcomes based on the percentage of SNPs genotyped or Het/Hom ratios, thresholds of percentage SNPs genotyped and Het/Hom ratio of approximately 85% and range within 0.3–0.45, respectively, were observed in order to predict a GEDmatch search outcome (Fig. 1). Three samples gave a Het/Hom ratio > 1, likely due to genotyping error causing false heterozygote SNP calls. All three samples also gave low SNP call rates

(< 52%). While increased Het/Hom ratio has been observed for cer- tain populations [23,24], all three samples were processed in du- plicate or triplicate with all associated repeats yielding a Het/Hom ratio < 1. Possible reasons considered for genotyping errors are in- sufficient DNA (input range 0.25 ng-0.04 ng), poor quality DNA (degradation index range 1.15–2.09) [25,26], artifacts introduced due to the deamination [27], contamination [28] or a combination of these.

To evaluate SNP call concordance, drop-out and drop-in, a pair- wise comparison was undertaken of all the ideal samples with an input of 80 ng or less to VIFM-01 (150 ng input). A positive correla- tion (Pearson correlation coefficient r = 0.94) was observed between concordance and a logarithmic increase in DNA input for samples with DNA input of 0.9 ng and under (Fig. 3). This relationship did not sustain for samples with a DNA input of 4 ng and over, as these samples reached close to maximum concordance. For drop-out, the comparisons showed that the number of drop-out decreased with an increasing DNA input; however, some drop-out (numbering in the 10 s) were observed in samples with 0.4 ng DNA input up to 80 ng DNA input (Fig. 4). Similarly for drop-in, the comparisons showed that the number of drop-ins decreased with an increasing DNA input; however, some drop-ins (numbering in the 10–100 s) were observed in samples with 0.6 ng DNA input up to 80 ng input DNA (Fig. 5).

3.2. Genetic Genealogy: Ideal sample – Dilution Series

The donor of the ideal sample identified as having Scottish, Dutch and Ashkenazi Jewish ancestry (Supplementary Fig. S1); with Dutch heritage going back two generations on the donor’s maternal side, and Australian/Scottish heritage on the paternal side. The donor’s ancestry was also determined using the admixture proportions provided by Eurogenes K13 modelling (Fig. S1), with North Atlantic (45%), Baltic (20%), West Mediterranean (15%), West Asian (10%) and East Mediterranean (7%) making up the bulk of the admixture. The DNA groups that make up the North Atlantic population include: Danish; French Basque; Irish; North Dutch; Norwegian; Orcadian; Southeast English; Southwest English and West Scottish. The ad- mixture prediction for this sample using the Eurogenes K13 mod- elling appeared to align with the donor’s reported ancestry.

The kits derived from the ideal sample – having incrementally less input DNA (from 80 ng down to 0.4 pg) – were compared to the kit obtained when 150 ng was used for the SNP analysis using the ‘one-to-one’ comparison tool in GEDmatch (Fig. S2). Kits down from 80 ng to 0.1 ng shared > 99% of their SNPs with VIFM-01 (150 ng) with greater than 550,000 SNPs used in the comparisons. For the two kits having 0.04 ng input (VIFM-06 and VIFM-32), they shared 81% and 95% respectively with VIFM-01 with 422,862 and 505,953 SNPs used in the comparisons. All the kits from 80 ng to 0.1 ng yielded Total Half-Match segments (HIR) > 3500 cM with a most- recent-common-ancestor (MRCA) of 1.0; while the two kits having 0.4 ng (VIFM-06 and VIFM-32) yielding HIR/MRCA values of 3027/1.1 and 3560/1.0 respectively. Kits with less than 0.02 ng failed to yield any results using the ‘one-to-one’ comparisons.

VIFM-01 (150 ng), when searched against all kits available on GEDmatch using the ‘one-to-many’ comparison, returned the ex- pected matches to other kits (five at the time of this comparison) held by the donor on GEDmatch (data not shown). Kits from 80 ng to 0.1 ng yield the same top five hits compared to VIFM-01 (150 ng) (Fig. S2) and while the top match was identical, the order of the other four matches varied slightly – with the top match having > 3500 shared cM and MRCA of 1. For the two 0.4 ng kits (VIFM-06 and VIFM-32), only the top match was the same as that of VIFM-01 (150 ng) with shared cM/MRCA values of 3161/1.1 and 3577/1.0 re- spectively. Kits with less than 0.02 ng returned what were con- sidered ‘junk’ matches (matches to known artificial junk kits present

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

in GEDmatch at the time) or no matches at all, likely as a result of having low SNP call rates (approximately 60% or less) in combination with an increase in genotyping errors due to insufficient and/or poor quality DNA available for typing.

3.3. Genetic Genealogy: Control samples

The donor ancestry for the C-samples was noted at the com- mencement of the project (data not shown), with the admixture proportions for the kits derived from the C-samples obtained using

the Eurogenes K13 modelling (Fig. S3). While the bulk of the ad- mixtures were composed of populations in the North Atlantic, Baltic and West Mediterranean, their proportions varied particularly be- tween C-sample 1 and the others – with C-sample 1 having a sig- nificant Amerindian contribution (27%). This is not surprising as the donor of this sample reported a South American and European an- cestry. The DNA groups that make up the Amerindian population include: Karitiana; Mayan; North Amerindian; and Pima. Hence the admixture prediction for this sample using the Eurogenes K13 modelling appeared to align with the donor’s reported ancestry. In

Fig. 1. Quality – An assessment of input amount and GEDmatch outcomes for non-degraded samples that yielded SNP data. Panel A: DNA input to Het/Hom ratio and GEDmatch outcomes. Panel B: DNA input to % SNP genotyped and GEDmatch outcomes.

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

Fig. 2. Quality – assessment of input amount and degradation on Het/Hom ratio and % SNPs genotyped for samples that yielded SNP data. Shown is DNA input and degradation to Het/Hom ratio and % SNPs genotyped.

Fig. 3. Quality – SNP call concordance for samples that yielded SNP data. Panel A: DNA input to the number of identical SNP calls between both samples. Panel B: DNA input to the number of non-identical SNP calls between both samples.

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

Fig. 4. Quality – SNP drop-out for samples that yielded SNP data. DNA input to drop-outs.

Fig. 5. Quality – SNP drop-in for samples that yielded SNP data. DNA input to drop-ins.

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

addition, C-sample 3 and C-sample 4, (which were derived from the same donor) generated almost identical admixture proportions (data not shown).

Of the ‘one-to-many’ comparisons, all kits for the C-samples generated matches on GEDmatch which could be further in- vestigated, these are detailed in Table 3. Kits for C-samples 3 and 4 (which originated from the same donor) produced the same match outcomes.

3.4. Genetic Genealogy: Casework Samples

Of the ‘one-to-many’ comparisons, only two of the eight case- work samples analysed generated matches on GEDmatch PRO which could be further investigated, in both instances the duplicate kits produced the same match outcomes. Of the remaining six cases, kits returned what were considered ‘junk’ matches or no matches at all, likely as a result of having poor SNP call rates and erroneous geno- typing due to insufficient DNA available for typing in some instances, or the DNA being degraded in others. This may indicate sample quality as the reason to matching ‘junk’ kits on the database (as different samples matched the same ‘junk’ kits) rather than not having any suitable matching kits on the database.

3.4. .1. Case 1 (VIFM-09 & VIFM-10) The admixture proportions for this case (Fig. S4) consisted of

populations in East Mediterranean (37%), West Mediterranean (19%), North Atlantic (17%), Baltic (9%), West Asian 8%) and Red Sea (6%) comprising the majority of the admixture. In addition, kits for VIFM- 09 and VIFM-10 generated almost identical admixture proportions (data not shown).

Of the ‘one-to-many’ comparison in GEDmatch Pro, Case 1 (VIFM-09 and VIFM-10) had the closest matches in the 80–500 range of total shared cM, indicative of second cousin (2 C) - third cousin (3 C) or first cousin once removed (1C1R) relationships (Table 3). The comparison performed on FamilyTreeDNA yielded a closer match (616 cM) placing this in the 550–1200 range indicative of first cousin (1 C)− 1C1R (data not shown).

3.4.2. . Case 8. (VIFM-42 & VIFM-43) The admixture proportions for this case (Fig. S4) consisted of

population in North Atlantic (49%), Baltic (22%), West Mediterranean (19%), West Asian (5%) and East Mediterranean (4%) comprising the majority of the admixture. In addition, kits for VIFM-42 and VIFM-43 gave almost identical admixture proportions (data not shown).

Of the ‘one-to-many’ comparison in GEDmatch PRO, Case 8 (VIFM-42 and VIFM-43) had a match in the 1200–2200 range in- dicative of half siblings, aunt, uncles, grand-parents/child relation- ships; as well as a further matches in the 80–500 range indicative of 2–3 C or 1C1R (Table 3).

3.5. Genetic Genealogy: Pristine Versus Compromised Samples

A comparison of GEDmatch and GEDmatch PRO outcomes for all samples that were successfully uploaded was undertaken. When evaluating Het/Hom ratio and the percentage of SNPs genotyped to GEDmatch or GEDmatch PRO outcomes, a Het/Hom ratio in the range of 0.3–0.45 was required to successfully match – even when the percentage of SNPs genotyped dropped below 90% (Fig. 6). Fur- thermore, the Het/Hom ratio appears to be a better predictor of match success rather than the percentage of SNPs genotyped (Fig. 7).

Furthermore, the impact that DNA input has on three types of genotyping errors (false allele call rate (homozygote); allele drop- out rate; and allele drop-in rate) was evaluated using the ideal sample series (see Fig. 8). When considering the three typing errors combined, it was noted that samples with an input of 0.02 ng or less had an observed genotyping error ranging from 19% to 63%; whereas samples with 0.1 ng or more had less than 1% genotyping errors, of which the majority were allele drop-ins. When looking at the 0.02 ng or less samples in more detail, allele drop-out is more prevalent for samples with 0.02 ng and 0.004 ng input, whereas the false allele rate at 32% and allele drop-out rate at 28%, are similar with the smallest DNA input of 0.0004 ng. For the duplicate samples with an input of 0.04 ng, each had genotyping errors of 18% and 5% respec- tively.

3.6. Bioinformatic Analysis: Casework Samples

In order to improve the search outcome for casework samples, bioinformatic analysis of the SNP data was undertaken to remove any non-concordant and no-data SNPs from the combined SNP data for those cases. The removal of all homozygote SNPs, or those not seen in at least two replicates was also performed. While the bioinformatic treatment improved the number of SNPs available for comparison for these casework samples (and thus the call rate) – as exemplified for case 4 (VIFM-15, VIFM-16, VIFM-38 and VIFM-39) (Table 4) – none of the kits generated from the upload of the treated data resulted in any match data (data not shown), as they all had Het/Hom ratios outside the range of 0.3–0.4 and failed this quality measure.

4. Discussion

While the operationalisation of FGG outside of the U.S. is limited, others have commenced evaluation of FGG for cold case investiga- tions and human identification purposes [15,29]. Tillmar et.al. 2021 [29], for example, described the successful use of FGG to solve a 16- year old double murder case in Sweden; demonstrating the utility of genetic genealogy databases for countries other than the U.S. As more countries consider the use of FGG, concerns have been raised regarding the ethical use of public databases for criminal

Table 3 One-to-many – control and casework samples. For control samples (VIFM-45, −46, −47 and −48) where a match list was returned from GEDmatch, and for casework samples (VIFM-09, −10, −42, and −43) where a match list was returned from GEDmatch PRO, the number of close matches based on cM is shown – with numbers capped to 500. Adapted from Thomson et. al. 2020 [34]

Number of close matches

Range cM VIFM-09 VIFM-10 VIFM-42 VIFM-43 VIFM-45 VIFM-46 VIFM-47 VIFM-48 Likely relationship

~ 3570 0 0 0 0 0 0 0 0 Parent or child 2200–3300 0 0 0 0 0 0 0 0 Full sibling 1200–2200 0 0 1 1 0 0 0 0 Half siblings, aunts, uncles,

grand-parents/children 550–1200 0 0 0 0 0 0 0 0 1–1 C1R 80–500 7 6 4 4 0 0 0 0 approximately 2–3 C or 1C1R 50–80 500 500 0 0 0 0 1 1 approximately 3–4 C 30–50 N/A N/A N/A N/A 5 4 10 9 approximately 4 C or more distant (could be closer)

N/A: not applicable

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

investigations, as well as the validity of methodology(ies) used to infer relationships [30–32]. Thus, case studies (using forensically relevant samples) to evaluate FGG are required to improve our awareness.

In order to better understand the impact of parameters such as DNA input and degradation levels as well as SNP call rates on the ability to deliver enough DNA data for a kit to yield matches when uploaded to a genealogy database, an ideal sample (at various input amounts), as well as control samples (varying degradation levels) were analysed. For a good quality DNA sample with a low de- gradation index (~ 1), as little as 0.1 ng was able to yield the expected match profile when performing the ‘one-to-many’ comparisons,

however, an input of 0.04 ng resulted in the identification of the same top match only. While inputs of 0.7 ng or greater were required to achieve the correct gender (as well as call rates above 0.97) this was not a requirement for success in yielding matches. Failure to correctly call the gender may be a result of not enough SNP markers on the X and/or Y chromosomes being typed to enable gender de- termination. However, this observation requires further investiga- tion. Hence, the ability to predict the gender of the donor from the SNP analysis was not a requirement to successfully yield matches when searched against GEDmatch, with samples returning ‘unknown’ for gender able to yield the expected matches. Samples with slightly worse degradation levels (ranging from 1.6 to 5.0) with

Fig. 7. GEDmatch and GEDmatch PRO outcomes. For all samples that were successfully uploaded to GEDmatch and GEDmatch Pro, an assessment of Het/Hom ratio and per- centage of SNPs genotyped to GEDmatch outcomes, following the removal of three outlier samples with Het/Hom ratios > 1.5.

Fig. 6. GEDmatch and GEDmatch PRO outcomes. For all samples that were successful uploaded to GEDmatch and GEDmatch Pro, an assessment of Het/Hom ratio and percentage of SNPs genotyped to GEDmatch outcomes.

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

inputs between 7 and 65 ng all generated matches when performing the ‘one-to-many’ comparisons. These results were encouraging as many of the samples received by the laboratory from UHR cases often yield low levels of DNA which are degraded. It was noted, however, that for samples with an input < 1 ng, degradation ap- peared to have an adverse impact on the Het/Hom ratio. For the ideal sample, it was observed that with DNA inputs less than 0.04 ng more than 5% genotyping errors were detected. Furthermore, the occur- rence of these genotyping errors coincided with the loss of a correct match list when searched in GEDmatch. The presence of 5% or more of genotyping errors had a detrimental impact on the matching outcome.

FGG has been used successfully in the identification of UHR cases, particularly in the U.S., leading to the reconciliation of long-term missing persons with unidentified deceased [7]. This approach is dependent on having the appropriate data set (size and composi- tion) in the genealogy database, such as GEDmatch and Family- TreeDNA, to ensure matches occur [33]. In order to evaluate the usefulness of FGG for the identification of UHR cases in an Australian

context, eight cases were selected for analysis. These cases typically represent the sample type (bone) and DNA yields and quality (low concentration and degraded) encountered for UHR in the laboratory. Furthermore, as it was expected that cases with European ancestry should yield results when searched on GEDmatch PRO, case selec- tion considered those cases which had BGA predictions available, with seven of the eight cases predicted to have European ancestry based on BGA analysis. Although Case 8 did not have any BGA pre- dictions available, mitochondrial haplotyping indicated European ancestry on the maternal side. The admixture predictions obtained using the Eurogenes K13 modelling appeared to be a good indicator of the ancestry for the known samples and could potentially be used as an indicator as to the likelihood of matches on the genealogy database for the unknown samples. However, further assessment of the accuracy of Eurogenes K13 modelling would be required.

For all the casework samples, DNA input amounts fell within the range of 0.1–14 ng, with all but one having degradation values of 0.6–3.6. Based on the results for the ideal sample, it was anticipated that most would return a match result using ‘one-to-many’

Fig. 8. GEDmatch and GEDmatch PRO outcomes. An assessment of genotyping error (%) as a function of input DNA for all samples that were successfully uploaded to GEDmatch and GEDmatch Pro.

Table 4 Bioinformatic – Case 4. Number of SNPs available for searching following three treatments: the removal (i) non-concordant and no-data SNPs; (ii) as (i) and any homozygote SNPs; and (iii) as (i) and any homozygote SNPs not seen in at least two of the replicates.

Bioinformatic Treatment # SNPs Call Rate Het/Hom # Non-concordant & no data SNPs removed

# Homozygote - removed

Nonea 509,495 0.71 0.08 (i) Non-concordant & no data SNPs removed 613,304 0.86 0.045 100,456 as (i) plus all homozygotes SNPS removed 29,077 0.04 N/A 100,456 584,705 as (i) plus all homozygotes SNPS not seen in at least two replicates

removed 557,722 0.78 0.055 100,456 56,060

a average values obtained from the four samples for Case 4

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

comparisons. However, this did not occur with only two of the eight cases returning a match result. Assessment of call rates showed four of the eight cases had call rates > 0.85, but of those only two had DNA input amounts > 12 ng, and it was these two cases which generated a match result. It should also be noted that these two cases yielded DNA of good quality (concentration and DI) which were derived from a bloodstain and fresh bone respectively. Based on the observation with the ideal sample, it was somewhat sur- prising that a case with an input amount (2 ng) and DI value (0.97) within values that had yielded a match result using the ideal sample returned a ‘junk’ match list suggesting that the SNP data was not of sufficient quality for matching. From the casework samples results, it could be said that call rate alone, following SNP analysis, is not a good indicator of success when conducting ‘one-to-many’ searches.

To improve the searchability of the casework samples that failed, bioinformatics was used to review and enrich the SNP data prior to uploading to the genealogy database. Of the four cases that failed to yield a match, the dataset for the corresponding samples were com- bined and cleaned (to remove any non-concordant SNPs, SNPs without any data, and in some instances Homozygote SNPs not seen in two or more replicates) before being searched against the database. This, however, did not improve the search outcome for these cases.

Consideration of the percentage of SNPs genotyped and Het/Hom ratio as indicators of sample quality, indicated that a threshold of 85% SNPs genotyped and Het/Hom ratio in the range of 0.3–0.45 was required in order to obtain a GEDmatch search outcome. This may be useful when predicting likely success or failure of matching (if re- lated individuals are in the database). Furthermore, the Het/Hom ratio value appeared to be a better predictor of match success, with the optimal value between 0.3 and 0.45. While these measures are only useful post WGA analysis, they may be helpful in evaluating search outcomes and where resources should be allocated.

When delving into the SNP data for the casework samples that failed to generate a match when searched against the genealogy database, it was noted that all displayed lower Het/Hom ratios (average 0.04) when compared to a good quality sample such as VIFM-25 (0.41). As noted above, for low quantity samples, the Het/ Hom ratio is adversely impacted by degradation. Given that these samples had DI values ranging from 1.15 to 3.6, this may have re- sulted in sub-optimal Het/Hom ratios for successful matching when combined with low input DNA. de Vries et.al. (2021), demonstrated that degradation dramatically impacted the likelihood of correctly identifying relationships [12].

Investigative leads provided by the two casework samples that yielded genealogy database matches require consideration of the next steps. This would involve the application of the genealogy process – building family trees and supporting links with historical documentation – in order to narrow down the potential identity of these individuals. The scope of this will be dependent on the com- plexity of the search required; with matches to 3rd- and 4th-cousins often requiring significant time and resources [34]. For the two casework samples that yielded matches, one had a match on GED- match PRO to a close relative (likely grandparent, aunt or uncle) which would be of greater assistance for the genealogy component compared to the other case which had more distant matches (likely 2–3 C or 1C1R). The use of more than one genealogy database may be required (such as GEDmatch PRO and FamilyTreeDNA) to poten- tially identify the closest relatives possible, and thus improve the genealogy outcomes. FamilyTreeDNA has over 1.1 million records on its database [35], while GEDmatch has over 1.1 million users (1.4 million DNA profiles) of which over 285,000 have opted-in to LE matching with 83% of new users opting-in to LE matching [14], hence, searching more than one database is likely to improve the matching outcomes. This will be case-to-case dependent and the benefits will need to be weighed against the costs associated with uploads.

While the application of FGG in the context of identifications is used more often in the U.S., with new identifications reported reg- ularly, this is not the case in Australia. Consideration of the devel- opment of best-practices is required to instil a level of confidence in the methodologies utilised in FGG, protect the genetic privacy of the donors and adhere to legislative requirements [36–40]. Although inroads have been made towards a better understanding of the statistical framework and methods used to infer distant relationship from dense SNP data [14,33], further work is needed to validate and develop guidelines for the application of FGG as is required for standard forensic DNA analysis [41] and to better understand the complexities of often compromised casework samples.

In Australia, consideration would need to be given as to the fra- mework under which any findings arising from an FGG investigation would be reported to the relevant authorities [16], and the need to undertake conventional forensic analysis to confirm any identity leads. Australians seemed to be engaged and interested in genealogy research; where in a month of monitoring, 9.4% of web traffic on GEDmatch.com originated from Australia [16]. The matching of two UHR cases to kits on the genealogy databases, has demonstrated the applicability of FGG (when DNA data of sufficient quantity and quality is available) in the development of investigative leads to assist in the identification of Jane and John Doe cases in Australia.

5. Conclusions

Although based on a limited sample size, this pilot study pro- vided useful insights into sample requirements for the successful matching of SNP data obtained using a WGA approach for casework samples. At a minimum, consideration needs to be given to the impact that input amount (> 1 ng for good quality samples) and sample degradation will have on the Het/Hom ratio; which was shown to be a good indicator of a match outcome. Based on the samples in this study, Het/Hom ratio is a useful indicator in addition to the other values (input amount, call rate etc) regarding reliability of results. Given the often-limited availability of compromised (low quantity/quality) casework samples for FGG analysis, the use of whole genome sequencing (WGS) instead of WGA should be con- sidered. A detailed evaluation of emerging FGG applications – such as Verogen’s Kintelligence – as well as other WGA (other than that used in this study) and WGS options, would be of immense value to the forensic community. Further work is currently being undertaken with 96 samples on two other WGAs to enable a more detailed as- sessment of suitable quality metrics for low quantity and quality samples.

CRediT authorship contribution statement

All persons who meet authorship criteria are listed as authors, with all authors participating in the concept, design, analysis, and writing and/or revision of the manuscript.

Declaration of Competing Interest

No potential conflicts of interest are reported by the authors.

Acknowledgements

The authors greatly acknowledge the support of the Victorian Institute of Forensic Medicine, and collaborative nature in which the Institute, Victoria Police and Coroners Court of Victoria work to- gether to identify unknown deceased persons. The technical ex- pertise of the Molecular Biology Laboratory was instrumental to the outcomes described in this publication. In addition, sincere thanks to other members of the Australian Forensic Genetic Genealogy

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242

Collaboration (Dr Nathan Scudder (Australian Federal Police), Dr Jennifer Raymond and Alison Sears (New South Wales Police)) for insightful discussions regarding FGG. The project was funded by internal operating budget of the Victorian Institute of Forensic Medicine. There are no external grants

Appendix A. Supporting information

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.forsciint.2022.111242.

References

[1] D.L. Ranson, L. Bugeja, Medicolegal Death Investgation: Coroner and Forensic Pahtology Functions and Processes in Victoria, Australia, Acad. Forensic Pathol. 7 (4) (2017) 567–581.

[2] D. Kennett, Using genetic genealogy databases in missing persons cases and to develop suspect leads in violent crimes, Forensic Sci. Int 301 (2019) 107–117.

[3] Y. Erlich, et al., Identity inference of genomic data using long-range familial searches, Science 362 (6415) (2018) 690–694.

[4] E.M. Greytak, C. Moore, S.L. Armentrout, Genetic genealogy for cold case and active investigations, Forensic Sci. Int 299 (2019) 103–113.

[5] S.H. Katsanis, Pedigrees and perpetrators: uses of DNA and genealogy in forensic investigations, Annu Rev. Genom. Hum. Genet 21 (2020) 535–564.

[6] C. Phillips, The Golden State Killer investigation and the nascent field of forensic genealogy, Forensic Sci. Int Genet 36 (2018) 186–188.

[7] DNA Doe Project . [cited 2021 30 April]; Available from: https://dnadoeproject.org/. [8] Buckskin girl : success story. [cited 2021 30 April]; Available from: https://

dnadoeproject.org/case/buckskin-girl/. [9] Snapshot Genetic Genealogy. [cited 2021 30 April]; Available from: https://

snapshot.parabon-nanolabs.com/genealogy. [10] Forensic Genealogy Service . [cited 2021 30 April]; Available from: https://www.

bodetech.com/FGS. [11] Justice through Genomics . [cited 2021 30 April]; Available from: https://othram.com/. [12] J.Hd Vries, et al., Impact of SNP microarray analysis of compromised DNA on

kinship classification success in the context of investigative genetic genealogy, Forensic Sci. Int Genet. 56 (2022) 102625.

[13] J. Ge, B. Budowle, Forensic investigation approaches of searching relatives in DNA databases, J. Forensic Sci. 66 (2) (2021) 430–443.

[14] D. Kling, et al., Investigative genetic genealogy: Current methods, knowledge and practice, Forensic Sci. Int Genet 52 (2021) 102474.

[15] A. Tillmar, et al., Whole-genome sequencing of human remains to enable gen- ealogy DNA database searches - A case report, Forensic Sci. Int Genet 46 (2020) 102233.

[16] N. Scudder, et al., Operationalising forensic genetic genealogy in an Australian context, Forensic Sci. Int 316 (2020) 110543.

[17] A. Schlenker, DNA extraction from skeletal remains: a comparison of four techniques with a view to improve DNA typing at the Victorian Institute of Forensic Medicine, in School of Biological Sciences, Monash Univ.: Melb., Aust. (2010) 104.

[18] Quantifiler ™ HP and Trio DNA Quantification Kits USER GUIDE. 12th October 2021]; Available from: https://tools.thermofisher.com/content/sfs/manuals/4485354.pdf.

[19] Infinium OmniExpress -24 v1.3 BeadChip. [cited 2021 30 April]; Available from: https://www.illumina.com/content/dam/illumina-marketing/documents/ products/datasheets/datasheet_human_omni_express.pdf.

[20] GEDmatch.com Terms of Service and Privacy Policy. Available from: https:// www.gedmatch.com/terms-of-service-privacy-policy.

[21] Updated Eurogenes K13 now at GEDmatch. Eurogenes Genetic Ancestry Project. [cited 2021 30 April]; Available from: http://bga101.blogspot.com/2013/11/ updated-eurogenes-k13-at-gedmatch.html.

[22] Forensic Testing & Genealogy Services. [cited 2021 19 October]; Available from: https://genebygene.com/forensics/.

[23] J. Wang, et al., Genome measures used for quality control are dependent on gene function and ancestry, Bioinformatics 31 (3) (2015) 318–323.

[24] D. Wu, et al., Large-scale whole-genome sequencing of three diverse asian po- pulations in Singapore, Cell 179 (3) (2019) 736–749 e15.

[25] J. Hasmats, et al., Assessment of whole genome amplification for sequence capture and massively parallel sequencing, PLoS One 9 (1) (2014) e84785.

[26] R. Do, et al., Exome sequencing identifies rare LDLR and APOA5 alleles conferring risk for myocardial infarction, Nature 518 (7537) (2015) 102–106.

[27] M. Hofreiter, et al., DNA sequences from multiple amplifications reveal artifacts induced by cytosine deamination in ancient DNA, Nucleic Acids Res 29 (23) (2001) 4793–4799.

[28] G. Jun, et al., Detecting and estimating contamination of human DNA samples in sequencing and array-based genotype data, Am. J. Hum. Genet 91 (5) (2012) 839–848.

[29] A. Tillmar, et al., Getting the conclusive lead with investigative genetic gen- ealogy - a successful case study of a 16 year old double murder in Sweden, Forensic Sci. Int Genet 53 (2021) 102525.

[30] C.J. Guerrini, et al., Should police have access to genetic genealogy databases? Capturing the Golden State Killer and other criminals using a controversial new forensic technique, PLoS Biol. 16 (10) (2018) e2006906.

[31] Syndercombe Court , D., Forensic genealogy: Some serious concerns. Forensic Sci Int Genet, 2018. 36: p. 203–204.

[32] A. Amorim, N. Pinto, Big data in forensic genetics, Forensic Sci. Int Genet 37 (2018) 102–105.

[33] D. Kling, A. Tillmar, Forensic genealogy-a comparison of methods to infer distant relationships based on dense SNP data, Forensic Sci. Int Genet 42 (2019) 113–124.

[34] J. Thomson, et al., An empirical investigation into the effectiveness of genetic genealogy to identify individuals in the UK, Forensic Sci. Int Genet 46 (2020) 102263.

[35] About The FamilyTreeDNA Database. [cited 2022 20 January]; Available from: https://www.familytreedna.com/why-ftdna.

[36] T.F. Callaghan, Responsible genetic genealogy, Science 366 (6462) (2019) 155. [37] N. Ram, J.L. Roberts, Forensic genealogy and the power of defaults, Nat.

Biotechnol. 37 (7) (2019) 707–708. [38] E. Murphy, Law and policy oversight of familial searches in recreational gen-

ealogy databases, Forensic Sci. Int 292 (2018) e5–e9. [39] G. Samuel, D. Kennett, The impact of investigative genetic genealogy: percep-

tions of UK professional and public stakeholders, Forensic Sci. Int Genet 48 (2020) 102366.

[40] R.A. Wickenheiser, Forensic genealogy, bioethics and the Golden State Killer case, Forensic Sci. Int Synerg 1 (2019) 114–125.

[41] Methods, S.W.G.o.D.A. Overview of investigative genetic genealogy. 2020 18 February 2020 [cited 2021 30 April]; Available from: https://1ecb9588-ea6f- 4feb-971a-73265dbf079c.filesusr.com/ugd/4344b0_ 6cc9e7c82ccc4fc0b5d10217af64e31b.pdf.

A. Davawala, A. Stock, M. Spiden et al. Forensic Science International 334 (2022) 111242