- Email: [email protected]
Efficacy of extended kinship analyses utilizing commercial STR kit in establishing personal identification
Legal Medicine 13 (2011) 12–15
Contents lists available at ScienceDirect
Legal Medicine journal homepage: www.elsevier.com/locate/legalmed
Efficacy of extended kinship analyses utilizing commercial STR kit in establishing personal identification Koichi Yoshida a, Kazuhiro Yayama a, Atsushi Hatanaka a, Keiji Tamaki b,⇑ a b
Forensic Science Laboratory, Kyoto Prefectural Police Headquarters, Kyoto 602-8550, Japan Department of Forensic Medicine and Molecular Pathology, Kyoto University, Graduate School of Medicine, Kyoto 606-8501, Japan
a r t i c l e
i n f o
Article history: Received 20 July 2010 Received in revised form 31 August 2010 Accepted 1 September 2010 Available online 16 October 2010 Keywords: Individual identification DNA typing STR Likelihood ratio Kinship
a b s t r a c t Unprecedented fidelity and specificity have afforded DNA testing its long reigning status as the gold standard for establishing personal identification. While the method itself is flawless, forensic experts have undoubtedly stumbled across challenging cases in which no reference samples for an unknown person (UP) are available for comparison. In such cases, experts often must resort to an assortment of kinship analyses—primarily those involving alleged parents or children of a UP—to establish personal identification. The present study derives likelihood ratio (LR) distributions from an extensive series of kinship simulations and places actual data, obtained from 120 cases in which personal identification of a UP was established via kinship analyses, to a comprehensive comparison in order to evaluate the efficacy of kinship assessments in establishing personal identification. A commercially available AmpFlSTR Identifiler kit was used to obtain DNA profiles. UP DNAs were extracted and isolated from fingernail (n = 87), cardiac blood (24), carpal bone (7) and tooth (2). Buccal cells were procured from alleged kin (AK) for subsequent kinship analyses. In 72 cases 1–3 alleged children were available for comparison; in 46 cases, one or both alleged parents were available; and in the final 2 cases (involving a pair of bodies discovered together in a dwelling), their alleged children were typed for comparison. For each case a LR was calculated based on the DNA typing results. Interestingly, we found that the median LR observed in the actual cases virtually mirrored those of the simulations. With exception to 2 cases in which a silent allele was observed at D19S433, biological relatives showed a LR greater than 100 and in these cases, kinship between the UP and AK were further supported by additional forms of evidence. We show here that in the vast majority of identification cases where direct reference samples are unavailable for a UP, kinship analyses referring to alleged parents/children and using 15 standard loci is more than capable of establishing the identification of a UP. However, discretion should be advised for silent alleles which—albeit rare—are known to occur at loci such as D19S433, along with other mutations which could render a deceivingly reduced LR. Ó 2010 Elsevier Ireland Ltd. All rights reserved.
1. Introduction In cases where the issue of personal identification lays in dispute, forensic experts confidently rely upon irrefutable DNA evidence to settle the case. In the Japanese forensic community the most widely applied system to establish a DNA profile is the AmpFlSTR Identifiler kit (Applied Biosystems). This is a multiplex PCR system that simultaneously amplifies 15 STR loci for a comprehensive DNA profile and also allows for sex determination by concomitantly targeting the amelogenin gene. Based on data from a previously published report on allele frequencies in the Japanese population [1], if a DNA profile consisting of all 15 AmpFlSTR is available, the average probability of a random match is 1.37 10 19 (Tamaki et al., unpublished data). Therefore, with the exception of monozygotic twins, this and similar ⇑ Corresponding author. Tel.: +81 75 753 4472; fax: +81 75 761 9591. E-mail address: [email protected] (K. Tamaki). 1344-6223/$ - see front matter Ó 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.legalmed.2010.09.001
systems possess an absolute power of discrimination. Another massive advantage of these systems is their inherent sensitivity which makes DNA profiling possibility even with limited amounts (e.g. subnanograms) of viable DNA to work with. The fully automated system which integrates capillary electrophoresis with computer-assisted STR genotyping not only makes reproducibility of the genotyping results possible, but also introduces the crucial element of objectivity into the overall analysis. Under ideal circumstances, a forensic DNA evaluation on an unidentified person will have available some sort of personal item (such as a razor or toothbrush) that can be used as a reference sample. This is the preferred method of engagement, largely because the analysis is untainted by complications such as the need to consider mutations in kinship analyses where multiple individuals must be assessed. Not only that but the likelihood ratio (LR)—which is the reciprocal of the random match probability—will exponentially increase in comparison to kinship analyses, and this will naturally
13
K. Yoshida et al. / Legal Medicine 13 (2011) 12–15
strengthen the link between the UP and the reference material. Unfortunately it’s not always the case that reference materials will be readily available, and often enough diverse circumstantial impediments like a fire extinguish even the most remote possibilities of retrieving any sort of reference materials. In such cases forensic experts must pursue other avenues of comparison such as kinship analyses in order to determine the identity of an unknown person (UP). In the present study, we present data obtained from 120 actual cases in which the identity of a UP was determined via kinship analyses. Furthermore we put the results obtained from the actual cases against a comparison to those obtained in a series of simulations in order to assess the veracity of kinship analyses in the identification of unidentified persons. 2. Materials and methods 2.1. Casework criteria This study was limited to 120 personal identification cases which resorted to kinship assessments between November 2006 and May 2009. In 24 cases where the postmortem interval (PI) was relatively short and the body was in appropriate condition, DNA was extracted and isolated from cardiac blood. In 87 cases with an extended PI, fingernails were adopted as the source for genomic DNA. In the remaining cases, the bodies of the UP were skeletonized, forcing the authors to pursue DNA isolation/extraction from teeth in 2 cases and carpal bones in 7 cases. For each case, buccal cells were procured from alleged kin (AK) for subsequent kinship analysis. 2.2. STR typing DNA was extracted from nail, blood, or bone samples procured from the deceased using an EZ1 DNA Investigator kit (Qiagen). For the family of a suspected individual, a Buccal DNA Collector (Bode Technology, Washington) was used to obtain buccal cells for subsequent DNA extraction using the BioRobotÒ EZ1 (Qiagen) found in the EZ1 DNA Investigator kit. Extraction was performed according to standard protocol. Sixteen loci, including amelogenin, were amplified from 0.75–1.0 ng of DNA. PCR conditions were programmed as instructed in the user manual. PCR products were then analyzed by an ABI 3130xl Genetic Analyzer (Applied Biosystems) and the genotypes were determined by Gene Mapper ID v3.2 software (Applied Biosystems).
Table 1 Classification of 120 actual cases. No. of Ups
AK’s alleged relationship
No. of AK
Designation
n
1
Child(ren)
1 2 3 1 2 1
U-C1 U-C2 U-C3 U-P1 U-P2 UU-C1
63a 6a 3 39b 7 2
Parent(s) 2
Child
a
In one case, an alleged spouse was introduced into the kinship assessment as an additional AK. b In one case, an alleged sibling introduced into the kinship assessment as an additional AK.
The varying degrees of kinship between a UP and AK we encountered in our 120 actual cases is classified and presented in Table 1. 1. Situations where a UP’s alleged child (AC) is available for comparison. (i) 1 AC available (U-C1); (ii) 2 AC available (U-C2); (iii) 3 AC available (U-C3). (Note: Sibship is established amongst AC in scenarios U-C2 and U-C3.) 2. Situations where a UP’s alleged parent (AP) is available for comparison. (i) 1 AP available (U-P1). In lieu of the AC being replaced by the AP, this scenario is equivalent to scenario 1(i) (i.e. U-C1). Therefore, hypothetically the LR value will be identical to scenario 1(i) if the genotypes for AP and AC are the same. (ii) 2 APs available (U-P2). Aside from popular misconception, the LR calculation here differs from the standard trio analysis applied to paternity tests because the relationship between a UP and the alleged mother has not yet been established. Consequently, the equation for calculating the LR also differs. 3. Situation in which there are 2 Ups in one case. There are two cases in which two UPs assumed to be husband and wife were discovered together. In these cases (UU-C1), the method for estimating the LR is exactly the same as in U-P2, but with the AK and UP being switched around (if an AC is available for comparison, the UP can be substituted for the AK—and vice versa—making the situation similar to that of U-P2). Therefore, the equation in 2(ii) can be applied here as well.
2.3. Kinship evaluation
3. Results and discussion
Determination of kinship was based on the LR of the probability of the UP and AKs respective DNA profiles occurring if the two parties were related, to the probability of the DNA profiles occurring if the UP and AK were unrelated. For each situational kinship evaluation posted below the genotypes were obtained for the UP and AK. Following an initial LR calculation for each locus, a combined LR was derived by combining the likelihood ratios of the individual 15 STR loci in each DNA profile. This LR was utilized in making the final assessment of kinship between a UP and AK. We have examined the efficacy of sibship analyses based on AmpFlSTR Identifiler loci in establishing personal identification in a previous study [2]. In addition to sibship analyses, we conducted an extensive series of 20,000 simulations encompassing a broad range of kinship scenarios in order to devise an expected LR distribution. The results obtained from these simulations were compared against the results obtained from actual casework. In order to ensure the quality and accuracy of our LR estimates, the results of our calculations were crossed-checked on DNA-VIEW software.
3.1. LR in cases where AK is assumed to be a child (U-C1), or a parent (U-P1), of UP We came across 63 cases in which the AK available for comparison was a single child (U-C1). In the vast majority of these cases (62/63) alleles were shared at all 15 STR loci thus resulting in no inconsistencies that the UP and AK were related. In one unique case, the AK (alleged son) and a male UP showed differing homozygous genotypes at D19S433, and therefore kinship was initially excluded. However, subsequent testing via Y-STR revealed a matching haplotype, which was later corroborated by further evidence and therefore we reversed our initial exclusionary position and determined that the AK and UP were related. Given these results, along with the fact that only D19S433 displayed an inconsistency with the original STR analysis, we assume that the AK and UP are in actuality heterozygous at this locus. The heterozygosity at this locus was most likely masked by a shared mutation which resulted in a silent allele. A report published by Mizuno et al. [3] on the Japanese population
14
K. Yoshida et al. / Legal Medicine 13 (2011) 12–15
attributes this silent allele to one of the primer annealing sites being compromised by a single base substitution. In the two Japanese populations Mizuno examines, this silent allele was identified in approximately 2% of the population—which equates to being present in 1 in every 50–60 paternity cases. This type of apparent opposite ‘‘homozygosity” can lead to wrongful exclusions if not careful addressed. The same silent allele was also reported by Tsuji et al. [4]. In terms of calculation, this silent allele can be treated as a mutation (in our case with a more than modest likelihood of 0.001). Therefore in our silent allele case, the combined LR was 17.8. Not including this isolated incident, in the remaining 62 cases the minimum LR was 129, the maximum LR was 3.06 108 and the average Log (LR) = 4.76 (equivalent to an LR of 5.7 104). In 39 cases the available kin was a single alleged parent (U-P1). In 38 cases, the AK and UP shared alleles at all loci and kinship could not be excluded; however, much like the isolated incident detailed above, in one case contradictory homozygous peaks were again observed at D19S433 which resulted in an initial exclusionary judgment. Much like in the U-C1 case detailed above, we attribute this to be a result of a silent allele. As in the earlier incident, we calculated the LR for D19S433 as 0.001 and ended up with a combined LR of 7.21. Putting this case aside, the minimum LR was 944, the maximum LR was 5.58 108 and the average Log (LR) = 4.52 (equivalent to an LR of 33,000). The formula used to derive the LR for U-P1 and U-C1 is exactly the same. Consequently, there was no significant difference for LR distribution in the two data sets (p = 0.112, Mann–Whitney U-test) and the two scenarios were combined in Fig. 1. In the 102 cases fitting within this bracket, only 2 were nearly excluded on account of the silent allele at D19S433, but in the remaining cases kinship could not be excluded. The LR range was 129 to 5.58 108 and the average logarithm for the LR was 4.67 (equivalent to an LR of 46,300). According to our simulation based data using the same scenario, the maximum LR is 4.32 1011, the minimum LR is 16.4 and the median is 48,900 [2]. When we calculate the common algorithm Log (LR), the mean is 4.77 and the standard deviation (SD) is 1.21; however, the pattern of distribution deviates from a normal distribution. In comparison, we see that the LR distribution between the actual and simulated cases relatively the same (p = 0.197, Mann–Whitney U-test) which incidentally demonstrates the veracity of the simulations.
Fig. 1. Combined LR of 63 U-C1 (UP and one alleged child) and 39 U-P1 (UP and one alleged parent) cases. Cases are sequentially placed in order from smallest to largest LR values. The two cases displaying the lowest values are shown as a shaded bar. The low LR in these cases is attributed to a silent allele occurring at D19S433, which had to be calculated as a mutation (LR = 0.001) at this locus. The dashed line denotes an LR of 500 (Log (LR) = 2.70).
If we use a prior probability of 50% and adopt Hummel’s standard of 99.8% or greater for inclusion, this equates to an LR of 500 (more specifically 499). In reality Identifiler based paternity simulations show that 2.8% (569 cases) of cases in which the alleged father and child are actually related will show a LR less than 500 [2]. In actual casework presented to the authors, roughly 6% (6 cases) of cases show a LR less than 500 (minimum 129, maximum 482). On the other end of the spectrum, in 3 simulations (0.02%) involving an AF with no biological ties to an alleged child, alleles were coincidentally shared at all 15 loci and therefore paternity could not be excluded. The LRs in these cases were 26.2, 91.8 and 484, but it should be noted that the vast majority of simulations (99.98%) in which the AF had no biological relation to the alleged child, paternity was rightly excluded. This shows us that even if the LR is low, if there are no inconsistencies with the genotyping results, we have a tendency to lean towards an inclusionary bias. We gathered the results obtained from 100 actual paternity tests that were performed, and compiled the minimum and maximum LRs for each locus in Fig. 2. The locus which most frequently displays the highest LR is D18S51 (18.0%), and the locus most frequently displaying the lowest LR is CSF1PO. Together, these results demonstrate that the locus with the highest power of discrimination (PD) [1] tends to have the maximum LR value within the 15 loci examined, and conversely, the locus demonstrating the lowest PD tends to correspond to the minimum LR value observed in a case.
3.2. LR in cases where multiple AKs are UP’s children We encountered 6 actual cases where the two available AK were the UP’s alleged children. In these cases the maximum LR was 4.22 1010, the minimum LR was 3.29 106 and the average LR logarithm was 8.14, which is consistent with the results obtained from our simulations. We also had 3 cases in which there were 3 alleged children available for comparison. The LR in this scenario was extremely high (range: 2.69 108 through 1.98 1011), obviously resulting in inclusionary results which ended with the UP being successfully identified. In the simulations where 2 children having both parents in common are compared with a non-related UP, all the relationship was excluded. The average number of excluded loci is 7.8 out of 15 and only 0.03% (6/20,000) was excluded at one locus.
Fig. 2. Loci demonstrating maximum and minimum LR values (of 15 loci examined) in U-C1 (n = 62) and U-P1 (n = 38) cases in relation to the powers of discrimination (PD) at each locus in the Japanese population. The two cases suspected to include a silent allele at D19S433 are excluded from this graph. PD values are derived from a previous report published by Yoshida et al. [1].
K. Yoshida et al. / Legal Medicine 13 (2011) 12–15
3.3. LR when two AKs are UP’s parents In seven actual cases both parents were available for comparison as the AK. In these cases the maximum LR was 2.64 1014, the minimum LR was 3.58 109 and the average Log (LR) = 12.2 (equivalent to an LR of 1.43 1012). The simulation-based results were comparable with a maximum LR of 1.47 1020, the minimum LR was 5.91 106 and median 6.68 1011. 95% of the actual cases fit within the simulation envelope. In the simulation setting where the scenario was set to identify whether the AKs are the parents of the UP, we achieved a flawless exclusion ratio when the UP was in fact biologically unrelated to the AK. In these simulations, exclusionary results were obtained at 4 loci at least, with the average number of loci demonstrating exclusion being 11 per case. For all of the actual cases we encountered in which kinship were assessed between an AK pair and a UP, no inconsistencies were documented at any loci, thus resulting in the successful identification of a UP. Two cases were brought to our intention in which both involved 2 UPs (suspected to be husband and wife) were discovered together in a dwelling, and an alleged child was asked to cooperate in the kinship analysis (UU-C1). In short, no inconsistencies were noted at any loci in either case. Essentially, the method for calculating the LR in this case is exactly the same as in U-P2 but with the AK and UP being switched around. The LRs were 7.04 1010 and 5.21 1010, respectively. Additionally, it’s worth noting here that LRs were obtained for comparisons involving each individual UP and AK, which also resulted in high LRs ranging 3660–13,300. 3.4. LR in other kinship cases In one unique case the two AKs available for comparison were the UPs alleged sibling and mother. The comparison involving the alleged mother and sibling on one side, and the UP on the other resulting in a LR of 1.96 105 which is more than enough to rule that the parties are in fact biologically related. Interestingly, individual assessments between the alleged sibling and UP resulted in a LR of only 36.8, whereas the individual assessment between the alleged mother and UP turned out a LR of 7.84 105 which was 4 times greater than the combined assessment between both AKs and the UP. Another interesting case involved that in which an alleged child of a UP (U-C1) and the UPs alleged spouse were available for analysis, making a standard trio analysis possible (LR 1.08 106, Log (LR) = 6.03). The LR we observed in this case was similar to that of the average logarithm of the simulations (6.89). We encountered another similar case but this time involving two alleged children (U-C2) and an alleged spouse. All loci demonstrated results consistent with kinship, with exception to D2S1338. While all parties were heterozygous at this locus, one of the children displayed an allele which was one repeat unit shy of the other parties. Therefore we considered this a mutation and calculated the LR at this locus to be 0.001. Even with the minor setback the combined LR was high (3.65 107, Log (LR) = 6.03) and we succeeded in establishing the UP as the father of these children. Given a challenging set of circumstances where reference materials of a UP could not be obtained, we resorted to a series of kinship analyses and calculated the LR based on results obtained from 15 STR loci. Under present analytical practices, Hummel’s predicate
15
is widely applied as the standard for determination in paternity assessments [5]. However, this is rather problematic given that Hummel’s predicate is not tailored specifically for the DNA testing scenarios presented in this study. For example, Evett and Weir contend that an LR greater than 1000 ‘‘very strongly supports” a hypothesis in criminal cases [6]. Numerous inconsistencies appear in previously reported publications on Identifiler-based sibship evaluations, and this urges the necessity for reconsidering the LR standard for these particular evaluations. Further consideration should be advised for half-sibship assessments as the degree of biological relation is even lesser than standard sibship. Additionally, had the AK been a UP’s half-siblings, parent, or child, it is necessary to reconsider the limitations of determining kinship using the present STR kit [7–9]. As the present study demonstrates, STR kits examining 15 loci can be effective in more than just standard comparisons, but with the proper attention devoted to common pitfalls in forensic mathematics the same kit equally suffices as a means for establishing personal identification via kinship evaluations. This is a monumental but lesser know tool in the existing arsenal of analytical techniques available to forensic scientists, and is especially useful in cases such as fires where reference DNA of a UP cannot be obtained from personal belongings. With the cooperation of alleged kin, personal identification can be achieved via a variety of kinship analyses. In the actual cases presented in this paper the kinship analyses are relatively clear cut because the AK are parents or children of a UP, which implies that alleles be shared at all loci. If multiple loci displayed contradictory genotypes (i.e. a lack of shared alleles) between an AK and UP, exclusion would have been a relatively simple task. In conclusion, given its potential for high LR values, resorting to kinship analyses for personal identification purposes is a highly appropriate option. However, when conducting kinship assessments, experts must be mindful of mutations and silent alleles as we observed in 3 cases, for these can lead to erroneous exclusions. Acknowledgements The authors would like to thank Dr. Richard H. Kaszynski for kindly providing us with professional editorial services. References [1] Yoshida K, Takahashi K, Kasai K. Allele frequencies of 15 loci using AmpFlSTR Identifiler kit in Japanese population. J Forensic Sci 2005;50:718–9. [2] Tamaki K, Kaszynski RH, Yuan QH, Yoshida K, Okuno T, Tsuruyama T. Likelihood evaluation using 15 common short tandem repeat loci: a practical and simulated approach to establishing personal identification via sibling/parental assessments. Transfusion 2009;49:578–84. [3] Mizuno N, Kitayama T, Fujii K, Nakahara H, Yoshida K, Sekiguchi K, et al. A D19S433 primer binding site mutation and the frequency in Japanese of the silent allele it causes. J Forensic Sci 2008;53:1068–73. [4] Tsuji A, Ishiko A, Umehara T, Usumoto Y, Hikiji W, Kudo K, et al. A silent allele in the locus D19S433 contained within the AmpFlSTR Identifiler PCR Amplification Kit. Leg Med 2010;12:94–6. [5] Hummel, K. Biostatistical opinion of parentage, 1971 Gustav Fischer Verlag, Stuttgart. (written in German and English). [6] Evett IW, Weir BS. Interpreting DNA Evidence. Statistical Genetics for Forensic Scientists. Sunderland: Sinauer Associates; 1998. pp. 217–246. [7] Allen RW, Fu J, Reid TM, Baird M. Considerations for interpretation on STR results in cases of questioned half-sibship. Transfusion 2007;47:515–9. [8] Wenk RE, Chiafari FA. Distinguishing full siblings from half-siblings in limited pedigrees. Transfusion 2000;40:44–7. [9] Wenk RE, Gjertson DW, Chiafari FA, Houtz T. The specific power of parentage exclusion in a child’s blood relatives. Transfusion 2005;45:440–4.
Contents lists available at ScienceDirect
Legal Medicine journal homepage: www.elsevier.com/locate/legalmed
Efficacy of extended kinship analyses utilizing commercial STR kit in establishing personal identification Koichi Yoshida a, Kazuhiro Yayama a, Atsushi Hatanaka a, Keiji Tamaki b,⇑ a b
Forensic Science Laboratory, Kyoto Prefectural Police Headquarters, Kyoto 602-8550, Japan Department of Forensic Medicine and Molecular Pathology, Kyoto University, Graduate School of Medicine, Kyoto 606-8501, Japan
a r t i c l e
i n f o
Article history: Received 20 July 2010 Received in revised form 31 August 2010 Accepted 1 September 2010 Available online 16 October 2010 Keywords: Individual identification DNA typing STR Likelihood ratio Kinship
a b s t r a c t Unprecedented fidelity and specificity have afforded DNA testing its long reigning status as the gold standard for establishing personal identification. While the method itself is flawless, forensic experts have undoubtedly stumbled across challenging cases in which no reference samples for an unknown person (UP) are available for comparison. In such cases, experts often must resort to an assortment of kinship analyses—primarily those involving alleged parents or children of a UP—to establish personal identification. The present study derives likelihood ratio (LR) distributions from an extensive series of kinship simulations and places actual data, obtained from 120 cases in which personal identification of a UP was established via kinship analyses, to a comprehensive comparison in order to evaluate the efficacy of kinship assessments in establishing personal identification. A commercially available AmpFlSTR Identifiler kit was used to obtain DNA profiles. UP DNAs were extracted and isolated from fingernail (n = 87), cardiac blood (24), carpal bone (7) and tooth (2). Buccal cells were procured from alleged kin (AK) for subsequent kinship analyses. In 72 cases 1–3 alleged children were available for comparison; in 46 cases, one or both alleged parents were available; and in the final 2 cases (involving a pair of bodies discovered together in a dwelling), their alleged children were typed for comparison. For each case a LR was calculated based on the DNA typing results. Interestingly, we found that the median LR observed in the actual cases virtually mirrored those of the simulations. With exception to 2 cases in which a silent allele was observed at D19S433, biological relatives showed a LR greater than 100 and in these cases, kinship between the UP and AK were further supported by additional forms of evidence. We show here that in the vast majority of identification cases where direct reference samples are unavailable for a UP, kinship analyses referring to alleged parents/children and using 15 standard loci is more than capable of establishing the identification of a UP. However, discretion should be advised for silent alleles which—albeit rare—are known to occur at loci such as D19S433, along with other mutations which could render a deceivingly reduced LR. Ó 2010 Elsevier Ireland Ltd. All rights reserved.
1. Introduction In cases where the issue of personal identification lays in dispute, forensic experts confidently rely upon irrefutable DNA evidence to settle the case. In the Japanese forensic community the most widely applied system to establish a DNA profile is the AmpFlSTR Identifiler kit (Applied Biosystems). This is a multiplex PCR system that simultaneously amplifies 15 STR loci for a comprehensive DNA profile and also allows for sex determination by concomitantly targeting the amelogenin gene. Based on data from a previously published report on allele frequencies in the Japanese population [1], if a DNA profile consisting of all 15 AmpFlSTR is available, the average probability of a random match is 1.37 10 19 (Tamaki et al., unpublished data). Therefore, with the exception of monozygotic twins, this and similar ⇑ Corresponding author. Tel.: +81 75 753 4472; fax: +81 75 761 9591. E-mail address: [email protected] (K. Tamaki). 1344-6223/$ - see front matter Ó 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.legalmed.2010.09.001
systems possess an absolute power of discrimination. Another massive advantage of these systems is their inherent sensitivity which makes DNA profiling possibility even with limited amounts (e.g. subnanograms) of viable DNA to work with. The fully automated system which integrates capillary electrophoresis with computer-assisted STR genotyping not only makes reproducibility of the genotyping results possible, but also introduces the crucial element of objectivity into the overall analysis. Under ideal circumstances, a forensic DNA evaluation on an unidentified person will have available some sort of personal item (such as a razor or toothbrush) that can be used as a reference sample. This is the preferred method of engagement, largely because the analysis is untainted by complications such as the need to consider mutations in kinship analyses where multiple individuals must be assessed. Not only that but the likelihood ratio (LR)—which is the reciprocal of the random match probability—will exponentially increase in comparison to kinship analyses, and this will naturally
13
K. Yoshida et al. / Legal Medicine 13 (2011) 12–15
strengthen the link between the UP and the reference material. Unfortunately it’s not always the case that reference materials will be readily available, and often enough diverse circumstantial impediments like a fire extinguish even the most remote possibilities of retrieving any sort of reference materials. In such cases forensic experts must pursue other avenues of comparison such as kinship analyses in order to determine the identity of an unknown person (UP). In the present study, we present data obtained from 120 actual cases in which the identity of a UP was determined via kinship analyses. Furthermore we put the results obtained from the actual cases against a comparison to those obtained in a series of simulations in order to assess the veracity of kinship analyses in the identification of unidentified persons. 2. Materials and methods 2.1. Casework criteria This study was limited to 120 personal identification cases which resorted to kinship assessments between November 2006 and May 2009. In 24 cases where the postmortem interval (PI) was relatively short and the body was in appropriate condition, DNA was extracted and isolated from cardiac blood. In 87 cases with an extended PI, fingernails were adopted as the source for genomic DNA. In the remaining cases, the bodies of the UP were skeletonized, forcing the authors to pursue DNA isolation/extraction from teeth in 2 cases and carpal bones in 7 cases. For each case, buccal cells were procured from alleged kin (AK) for subsequent kinship analysis. 2.2. STR typing DNA was extracted from nail, blood, or bone samples procured from the deceased using an EZ1 DNA Investigator kit (Qiagen). For the family of a suspected individual, a Buccal DNA Collector (Bode Technology, Washington) was used to obtain buccal cells for subsequent DNA extraction using the BioRobotÒ EZ1 (Qiagen) found in the EZ1 DNA Investigator kit. Extraction was performed according to standard protocol. Sixteen loci, including amelogenin, were amplified from 0.75–1.0 ng of DNA. PCR conditions were programmed as instructed in the user manual. PCR products were then analyzed by an ABI 3130xl Genetic Analyzer (Applied Biosystems) and the genotypes were determined by Gene Mapper ID v3.2 software (Applied Biosystems).
Table 1 Classification of 120 actual cases. No. of Ups
AK’s alleged relationship
No. of AK
Designation
n
1
Child(ren)
1 2 3 1 2 1
U-C1 U-C2 U-C3 U-P1 U-P2 UU-C1
63a 6a 3 39b 7 2
Parent(s) 2
Child
a
In one case, an alleged spouse was introduced into the kinship assessment as an additional AK. b In one case, an alleged sibling introduced into the kinship assessment as an additional AK.
The varying degrees of kinship between a UP and AK we encountered in our 120 actual cases is classified and presented in Table 1. 1. Situations where a UP’s alleged child (AC) is available for comparison. (i) 1 AC available (U-C1); (ii) 2 AC available (U-C2); (iii) 3 AC available (U-C3). (Note: Sibship is established amongst AC in scenarios U-C2 and U-C3.) 2. Situations where a UP’s alleged parent (AP) is available for comparison. (i) 1 AP available (U-P1). In lieu of the AC being replaced by the AP, this scenario is equivalent to scenario 1(i) (i.e. U-C1). Therefore, hypothetically the LR value will be identical to scenario 1(i) if the genotypes for AP and AC are the same. (ii) 2 APs available (U-P2). Aside from popular misconception, the LR calculation here differs from the standard trio analysis applied to paternity tests because the relationship between a UP and the alleged mother has not yet been established. Consequently, the equation for calculating the LR also differs. 3. Situation in which there are 2 Ups in one case. There are two cases in which two UPs assumed to be husband and wife were discovered together. In these cases (UU-C1), the method for estimating the LR is exactly the same as in U-P2, but with the AK and UP being switched around (if an AC is available for comparison, the UP can be substituted for the AK—and vice versa—making the situation similar to that of U-P2). Therefore, the equation in 2(ii) can be applied here as well.
2.3. Kinship evaluation
3. Results and discussion
Determination of kinship was based on the LR of the probability of the UP and AKs respective DNA profiles occurring if the two parties were related, to the probability of the DNA profiles occurring if the UP and AK were unrelated. For each situational kinship evaluation posted below the genotypes were obtained for the UP and AK. Following an initial LR calculation for each locus, a combined LR was derived by combining the likelihood ratios of the individual 15 STR loci in each DNA profile. This LR was utilized in making the final assessment of kinship between a UP and AK. We have examined the efficacy of sibship analyses based on AmpFlSTR Identifiler loci in establishing personal identification in a previous study [2]. In addition to sibship analyses, we conducted an extensive series of 20,000 simulations encompassing a broad range of kinship scenarios in order to devise an expected LR distribution. The results obtained from these simulations were compared against the results obtained from actual casework. In order to ensure the quality and accuracy of our LR estimates, the results of our calculations were crossed-checked on DNA-VIEW software.
3.1. LR in cases where AK is assumed to be a child (U-C1), or a parent (U-P1), of UP We came across 63 cases in which the AK available for comparison was a single child (U-C1). In the vast majority of these cases (62/63) alleles were shared at all 15 STR loci thus resulting in no inconsistencies that the UP and AK were related. In one unique case, the AK (alleged son) and a male UP showed differing homozygous genotypes at D19S433, and therefore kinship was initially excluded. However, subsequent testing via Y-STR revealed a matching haplotype, which was later corroborated by further evidence and therefore we reversed our initial exclusionary position and determined that the AK and UP were related. Given these results, along with the fact that only D19S433 displayed an inconsistency with the original STR analysis, we assume that the AK and UP are in actuality heterozygous at this locus. The heterozygosity at this locus was most likely masked by a shared mutation which resulted in a silent allele. A report published by Mizuno et al. [3] on the Japanese population
14
K. Yoshida et al. / Legal Medicine 13 (2011) 12–15
attributes this silent allele to one of the primer annealing sites being compromised by a single base substitution. In the two Japanese populations Mizuno examines, this silent allele was identified in approximately 2% of the population—which equates to being present in 1 in every 50–60 paternity cases. This type of apparent opposite ‘‘homozygosity” can lead to wrongful exclusions if not careful addressed. The same silent allele was also reported by Tsuji et al. [4]. In terms of calculation, this silent allele can be treated as a mutation (in our case with a more than modest likelihood of 0.001). Therefore in our silent allele case, the combined LR was 17.8. Not including this isolated incident, in the remaining 62 cases the minimum LR was 129, the maximum LR was 3.06 108 and the average Log (LR) = 4.76 (equivalent to an LR of 5.7 104). In 39 cases the available kin was a single alleged parent (U-P1). In 38 cases, the AK and UP shared alleles at all loci and kinship could not be excluded; however, much like the isolated incident detailed above, in one case contradictory homozygous peaks were again observed at D19S433 which resulted in an initial exclusionary judgment. Much like in the U-C1 case detailed above, we attribute this to be a result of a silent allele. As in the earlier incident, we calculated the LR for D19S433 as 0.001 and ended up with a combined LR of 7.21. Putting this case aside, the minimum LR was 944, the maximum LR was 5.58 108 and the average Log (LR) = 4.52 (equivalent to an LR of 33,000). The formula used to derive the LR for U-P1 and U-C1 is exactly the same. Consequently, there was no significant difference for LR distribution in the two data sets (p = 0.112, Mann–Whitney U-test) and the two scenarios were combined in Fig. 1. In the 102 cases fitting within this bracket, only 2 were nearly excluded on account of the silent allele at D19S433, but in the remaining cases kinship could not be excluded. The LR range was 129 to 5.58 108 and the average logarithm for the LR was 4.67 (equivalent to an LR of 46,300). According to our simulation based data using the same scenario, the maximum LR is 4.32 1011, the minimum LR is 16.4 and the median is 48,900 [2]. When we calculate the common algorithm Log (LR), the mean is 4.77 and the standard deviation (SD) is 1.21; however, the pattern of distribution deviates from a normal distribution. In comparison, we see that the LR distribution between the actual and simulated cases relatively the same (p = 0.197, Mann–Whitney U-test) which incidentally demonstrates the veracity of the simulations.
Fig. 1. Combined LR of 63 U-C1 (UP and one alleged child) and 39 U-P1 (UP and one alleged parent) cases. Cases are sequentially placed in order from smallest to largest LR values. The two cases displaying the lowest values are shown as a shaded bar. The low LR in these cases is attributed to a silent allele occurring at D19S433, which had to be calculated as a mutation (LR = 0.001) at this locus. The dashed line denotes an LR of 500 (Log (LR) = 2.70).
If we use a prior probability of 50% and adopt Hummel’s standard of 99.8% or greater for inclusion, this equates to an LR of 500 (more specifically 499). In reality Identifiler based paternity simulations show that 2.8% (569 cases) of cases in which the alleged father and child are actually related will show a LR less than 500 [2]. In actual casework presented to the authors, roughly 6% (6 cases) of cases show a LR less than 500 (minimum 129, maximum 482). On the other end of the spectrum, in 3 simulations (0.02%) involving an AF with no biological ties to an alleged child, alleles were coincidentally shared at all 15 loci and therefore paternity could not be excluded. The LRs in these cases were 26.2, 91.8 and 484, but it should be noted that the vast majority of simulations (99.98%) in which the AF had no biological relation to the alleged child, paternity was rightly excluded. This shows us that even if the LR is low, if there are no inconsistencies with the genotyping results, we have a tendency to lean towards an inclusionary bias. We gathered the results obtained from 100 actual paternity tests that were performed, and compiled the minimum and maximum LRs for each locus in Fig. 2. The locus which most frequently displays the highest LR is D18S51 (18.0%), and the locus most frequently displaying the lowest LR is CSF1PO. Together, these results demonstrate that the locus with the highest power of discrimination (PD) [1] tends to have the maximum LR value within the 15 loci examined, and conversely, the locus demonstrating the lowest PD tends to correspond to the minimum LR value observed in a case.
3.2. LR in cases where multiple AKs are UP’s children We encountered 6 actual cases where the two available AK were the UP’s alleged children. In these cases the maximum LR was 4.22 1010, the minimum LR was 3.29 106 and the average LR logarithm was 8.14, which is consistent with the results obtained from our simulations. We also had 3 cases in which there were 3 alleged children available for comparison. The LR in this scenario was extremely high (range: 2.69 108 through 1.98 1011), obviously resulting in inclusionary results which ended with the UP being successfully identified. In the simulations where 2 children having both parents in common are compared with a non-related UP, all the relationship was excluded. The average number of excluded loci is 7.8 out of 15 and only 0.03% (6/20,000) was excluded at one locus.
Fig. 2. Loci demonstrating maximum and minimum LR values (of 15 loci examined) in U-C1 (n = 62) and U-P1 (n = 38) cases in relation to the powers of discrimination (PD) at each locus in the Japanese population. The two cases suspected to include a silent allele at D19S433 are excluded from this graph. PD values are derived from a previous report published by Yoshida et al. [1].
K. Yoshida et al. / Legal Medicine 13 (2011) 12–15
3.3. LR when two AKs are UP’s parents In seven actual cases both parents were available for comparison as the AK. In these cases the maximum LR was 2.64 1014, the minimum LR was 3.58 109 and the average Log (LR) = 12.2 (equivalent to an LR of 1.43 1012). The simulation-based results were comparable with a maximum LR of 1.47 1020, the minimum LR was 5.91 106 and median 6.68 1011. 95% of the actual cases fit within the simulation envelope. In the simulation setting where the scenario was set to identify whether the AKs are the parents of the UP, we achieved a flawless exclusion ratio when the UP was in fact biologically unrelated to the AK. In these simulations, exclusionary results were obtained at 4 loci at least, with the average number of loci demonstrating exclusion being 11 per case. For all of the actual cases we encountered in which kinship were assessed between an AK pair and a UP, no inconsistencies were documented at any loci, thus resulting in the successful identification of a UP. Two cases were brought to our intention in which both involved 2 UPs (suspected to be husband and wife) were discovered together in a dwelling, and an alleged child was asked to cooperate in the kinship analysis (UU-C1). In short, no inconsistencies were noted at any loci in either case. Essentially, the method for calculating the LR in this case is exactly the same as in U-P2 but with the AK and UP being switched around. The LRs were 7.04 1010 and 5.21 1010, respectively. Additionally, it’s worth noting here that LRs were obtained for comparisons involving each individual UP and AK, which also resulted in high LRs ranging 3660–13,300. 3.4. LR in other kinship cases In one unique case the two AKs available for comparison were the UPs alleged sibling and mother. The comparison involving the alleged mother and sibling on one side, and the UP on the other resulting in a LR of 1.96 105 which is more than enough to rule that the parties are in fact biologically related. Interestingly, individual assessments between the alleged sibling and UP resulted in a LR of only 36.8, whereas the individual assessment between the alleged mother and UP turned out a LR of 7.84 105 which was 4 times greater than the combined assessment between both AKs and the UP. Another interesting case involved that in which an alleged child of a UP (U-C1) and the UPs alleged spouse were available for analysis, making a standard trio analysis possible (LR 1.08 106, Log (LR) = 6.03). The LR we observed in this case was similar to that of the average logarithm of the simulations (6.89). We encountered another similar case but this time involving two alleged children (U-C2) and an alleged spouse. All loci demonstrated results consistent with kinship, with exception to D2S1338. While all parties were heterozygous at this locus, one of the children displayed an allele which was one repeat unit shy of the other parties. Therefore we considered this a mutation and calculated the LR at this locus to be 0.001. Even with the minor setback the combined LR was high (3.65 107, Log (LR) = 6.03) and we succeeded in establishing the UP as the father of these children. Given a challenging set of circumstances where reference materials of a UP could not be obtained, we resorted to a series of kinship analyses and calculated the LR based on results obtained from 15 STR loci. Under present analytical practices, Hummel’s predicate
15
is widely applied as the standard for determination in paternity assessments [5]. However, this is rather problematic given that Hummel’s predicate is not tailored specifically for the DNA testing scenarios presented in this study. For example, Evett and Weir contend that an LR greater than 1000 ‘‘very strongly supports” a hypothesis in criminal cases [6]. Numerous inconsistencies appear in previously reported publications on Identifiler-based sibship evaluations, and this urges the necessity for reconsidering the LR standard for these particular evaluations. Further consideration should be advised for half-sibship assessments as the degree of biological relation is even lesser than standard sibship. Additionally, had the AK been a UP’s half-siblings, parent, or child, it is necessary to reconsider the limitations of determining kinship using the present STR kit [7–9]. As the present study demonstrates, STR kits examining 15 loci can be effective in more than just standard comparisons, but with the proper attention devoted to common pitfalls in forensic mathematics the same kit equally suffices as a means for establishing personal identification via kinship evaluations. This is a monumental but lesser know tool in the existing arsenal of analytical techniques available to forensic scientists, and is especially useful in cases such as fires where reference DNA of a UP cannot be obtained from personal belongings. With the cooperation of alleged kin, personal identification can be achieved via a variety of kinship analyses. In the actual cases presented in this paper the kinship analyses are relatively clear cut because the AK are parents or children of a UP, which implies that alleles be shared at all loci. If multiple loci displayed contradictory genotypes (i.e. a lack of shared alleles) between an AK and UP, exclusion would have been a relatively simple task. In conclusion, given its potential for high LR values, resorting to kinship analyses for personal identification purposes is a highly appropriate option. However, when conducting kinship assessments, experts must be mindful of mutations and silent alleles as we observed in 3 cases, for these can lead to erroneous exclusions. Acknowledgements The authors would like to thank Dr. Richard H. Kaszynski for kindly providing us with professional editorial services. References [1] Yoshida K, Takahashi K, Kasai K. Allele frequencies of 15 loci using AmpFlSTR Identifiler kit in Japanese population. J Forensic Sci 2005;50:718–9. [2] Tamaki K, Kaszynski RH, Yuan QH, Yoshida K, Okuno T, Tsuruyama T. Likelihood evaluation using 15 common short tandem repeat loci: a practical and simulated approach to establishing personal identification via sibling/parental assessments. Transfusion 2009;49:578–84. [3] Mizuno N, Kitayama T, Fujii K, Nakahara H, Yoshida K, Sekiguchi K, et al. A D19S433 primer binding site mutation and the frequency in Japanese of the silent allele it causes. J Forensic Sci 2008;53:1068–73. [4] Tsuji A, Ishiko A, Umehara T, Usumoto Y, Hikiji W, Kudo K, et al. A silent allele in the locus D19S433 contained within the AmpFlSTR Identifiler PCR Amplification Kit. Leg Med 2010;12:94–6. [5] Hummel, K. Biostatistical opinion of parentage, 1971 Gustav Fischer Verlag, Stuttgart. (written in German and English). [6] Evett IW, Weir BS. Interpreting DNA Evidence. Statistical Genetics for Forensic Scientists. Sunderland: Sinauer Associates; 1998. pp. 217–246. [7] Allen RW, Fu J, Reid TM, Baird M. Considerations for interpretation on STR results in cases of questioned half-sibship. Transfusion 2007;47:515–9. [8] Wenk RE, Chiafari FA. Distinguishing full siblings from half-siblings in limited pedigrees. Transfusion 2000;40:44–7. [9] Wenk RE, Gjertson DW, Chiafari FA, Houtz T. The specific power of parentage exclusion in a child’s blood relatives. Transfusion 2005;45:440–4.