- Email: [email protected]
Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis
Legal Medicine xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Legal Medicine journal homepage: www.elsevier.com/locate/legalmed
Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis Tomonori Tamura a,b, Motoki Osawa a,⇑, Eriko Ochiai a, Takanori Suzuki b, Takashi Nakamura b a b
Department of Forensic Medicine, Tokai University School of Medicine, Shimokasuya 143, Isehara, Kanagawa 259-1193, Japan Scientific Crime Laboratory, Kanagawa Prefectural Police, Yamashitacho 155, Naka-ku, Yokohama 231-0023, Japan
a r t i c l e
i n f o
Article history: Received 24 September 2014 Received in revised form 23 March 2015 Accepted 23 March 2015 Available online xxxx Keywords: Kinship analysis Short tandem repeat Likelihood ratio PowerPlex Fusion GlobalFiler PowerPlex 21
a b s t r a c t The AmpFLSTR Identifiler Kit, comprising 15 autosomal short tandem repeat (STR) loci, is commonly employed in forensic practice for calculating match probabilities and parentage testing. The conventional system exhibits insufficient estimation for kinship analysis such as sibship testing because of shortness of examined loci. This study evaluated the power of the PowerPlex Fusion System, GlobalFiler Kit, and PowerPlex 21 System, which comprise more than 20 autosomal STR loci, to estimate pairwise blood relatedness (i.e., parent–child, full siblings, second-degree relatives, and first cousins). The genotypes of all 24 STR loci in 10,000 putative pedigrees were constructed by simulation. The likelihood ratio for each locus was calculated from joint probabilities for relatives and non-relatives. The combined likelihood ratio was calculated according to the product rule. The addition of STR loci improved separation between relatives and non-relatives. However, these systems were less effectively extended to the inference for first cousins. In conclusion, these advanced systems will be useful in forensic personal identification, especially in the evaluation of full siblings and second-degree relatives. Moreover, the additional loci may give rise to two major issues of more frequent mutational events and several pairs of linked loci on the same chromosome. Ó 2015 Published by Elsevier Ireland Ltd.
1. Introduction Pedigree analysis in forensic science plays a major role in the identification of missing persons in mass disasters and accidents. In cases in which the personal remains of the deceased are unavailable, indirect identification is often performed by confirmation of any kinship with a potential family member. In addition to parentage testing, DNA analysis for other relatives such as full siblings, second-degree relatives, and first cousins is sometimes required in Japan. The clarification of the relationship between two individuals is required in the most common situation. For instance, forensic investigators are asked questions such as if the alleged father is the true father of the child or not and whether two males are brothers or not. The Paternity Testing Commission of the International Society of Forensic Genetics (ISFG) recommends the likelihood ratio (LR) approach if the weight of genetic evidence is calculated in parentage testing [1], although other parameters Abbreviations: AABB, American Association of Blood Banks; ISFG, International Society of Forensic Genetics; STR, short tandem repeat; LR, likelihood ratio. ⇑ Corresponding author. Tel.: +81 463 93 1121; fax: +81 463 92 0284. E-mail address: [email protected] (M. Osawa).
are also utilized for parentage testing (e.g., exclusion probability and probability based on Bayes’ theorem). The LR approach can be extended to other relationships such as full and half siblings [2,3]. When the genotypes used are mutually independent, the combined LR is calculated by multiplying LRs according to the product rule. If necessary, the LR and combined LR are adjusted for the effects of subpopulation and linkage [4]. Parentage is inferred according to typical cutoffs recommended by the ISFG: 100 or 1000 [5]. In sibship testing, a cutoff of 6100 is usually acceptable according to the American Association of Blood Banks (AABB) [6]. However, there is no universal cutoff among laboratories. An LR of 500 [7], which is Hummel’s paternity criterion during the blood group typing, is preferred in Japan regardless of the type of kinship. Genetic polymorphisms of short tandem repeat (STR) play a critical role in personal identification. The AmpFLSTRÒ IdentifilerÒ PCR Amplification Kit (Applied Biosystems, Foster City, CA, USA), which is composed of 15 autosomal STR loci (13 common loci of CODIS, D2S1338, and D19S433) and the Amelogenin locus, has been adopted as the nationwide standard in Japan. This system is used for the match probability calculation as well as LR calculation in duo and trio parentage testing.
http://dx.doi.org/10.1016/j.legalmed.2015.03.005 1344-6223/Ó 2015 Published by Elsevier Ireland Ltd.
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
2
T. Tamura et al. / Legal Medicine xxx (2015) xxx–xxx
However, Tamaki et al. previously found that this kit occasionally has limitations when obtaining a significant LR in pairwise sibship testing [8,9]. This limitation can be overcome by analyzing additional autosomal STR loci [10]. There are three commercially available major multiplex STR systems that comprise more than 20 autosomal STR loci: the PowerPlexÒ Fusion System (Promega Corporation, Madison, WI, USA), which comprises 22 autosomal STR loci of 13 CODIS, D1S1656, D2S441, D2S1338, D10S1248, D12S391, D19S433, D22S1045, Penta D, and Penta E; the GlobalFilerÒ PCR Amplification Kit (Applied Biosystems), which comprises 21 autosomal STR loci of 13 CODIS, D1S1656, D2S441, D2S1338, SE33, D10S1248, D12S391, D19S433, and D22S1045; and the PowerPlexÒ 21 System (Promega Corporation), which comprises 20 autosomal STR loci of 13 CODIS, D1S1656, D2S1338, D6S1043, D12S391, D19S433, Penta D, and Penta E. These three systems include five to seven loci in addition to those in the Identifiler. This study evaluated the power of the PowerPlex Fusion, GlobalFiler, and PowerPlex 21 to solve different cases of pairwise kinship analysis using simulated genotype data. In addition, the analyses were performed with the maximum set of 24 autosomal STR loci included in the three advanced systems. Finally, we discussed the problems of mutation and linkage associated with the addition of examined loci. 2. Materials and methods 2.1. Simulated STR type data for putative pedigrees For the simulation study, 10,000 datasets of five different pairwise relations of parent–child, full siblings, second-degree relatives, first cousins and non-relatives were constructed under a putative pedigree chart, representatively showed in Fig. 1. Individuals #1 and #6 represent the sets as parent–child, #5 and #6 as full siblings, #5 and #7 as second-degree relatives, #7 and #8 as first cousins, and #3 and #4 as non-relatives. Second-degree relatives are two meioses away from a particular individual in a pedigree (i.e., grandparent–grandchild, uncle/aunt–nephew/niece, and half-siblings). First, a series of genotypes in individuals #1– #4 were constructed on the basis of the allele frequencies of the 24 STR loci in the Japanese population [11–13]; then, those in individuals #5–#8 were constructed by mating of those of #1– #4. Mutational events were not considered in this process. Uniform random numbers were produced using R version 3.1.0 [14], and the construction of STR type data in the putative pedigrees were conducted in Microsoft Excel (2010).
relatives (H1) vs. non-relatives (H2), and first cousins (H1) vs. nonrelatives (H2)—were investigated by simulations. The simulation was performed with 10,000 sets of STR types where hypothesis H1 was assumed to be true, and those where hypothesis H2 was assumed to be true. The LR for each locus was calculated by dividing the joint probability assuming the two individuals were relatives by that assuming the two individuals were non-relatives on the basis of the allele frequencies in the Japanese population. For the five sets of the 24 STR loci, PowerPlex Fusion, GlobalFiler, PowerPlex 21, and Identifiler, the combined LRs were calculated by multiplying LRs based on the product rule. The linkage and subpopulation were not considered in this calculation. The combined LR was calculated using Microsoft Excel (2010). 2.3. Distribution of combined LRs, sensitivity, and specificity The distribution of the combined LR was depicted as a density curve using R version 3.1.0. In addition, sensitivity (the probability of judging relatives correctly as relatives), specificity (the probability of judging non-relatives correctly as non-relatives), positive predictive value (the proportion of subjects correctly judged as relatives), and negative predictive value (the proportion of subjects correctly judged as non-relatives) were calculated for different cutoffs of the combined LR. The power of the 24 STR loci, PowerPlex Fusion, GlobalFiler, and PowerPlex 21 to solve cases in the above mentioned scenarios was evaluated by comparing the sensitivity, specificity, positive predictive value, and negative predictive value of these advanced systems with those of Identifiler. All values were calculated using Microsoft Excel (2010). This project was approved by the Ethics Committee of the Tokai University School of Medicine. 3. Results and discussion 3.1. Parent–child scenario For the five sets of 24 STR loci, PowerPlex Fusion, GlobalFiler, PowerPlex 21, and Identifiler, the distribution of the common logarithm of the combined LR observed in the putative parent–child
2.2. Combined LR Four different case scenarios—parent–child (H1) vs. nonrelatives (H2), full siblings (H1) vs. non-relatives (H2), second-degree
#3
#8
#1
#2
#5
#6
#4
#7
Fig. 1. Putative pedigree chart for the construction of pairwise kinship.
Fig. 2. Distribution of combined LR (common logarithm) in parent–child.
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
T. Tamura et al. / Legal Medicine xxx (2015) xxx–xxx
relation is indicated in Fig. 2. The maximum, minimum, and mean ± SD of the combined LR calculated in the parent–child are summarized in Table 1 and the sensitivity is presented in Table 2. The mean, variance, and maximum of the combined LR were raised with the increasing number of STR loci. The minimum combined LR did not depend on the number of STR loci, which was presumed to be attributable to the outliers or extreme cases of the simulation data as well as differences in STR loci included in the systems. The additional STR loci generally improved sensitivity. If a standard cutoff of the combined LR of 100 or 1000 [5] was adopted, even Identifiler showed sufficiently high sensitivity up to 0.9963 and 0.9461, respectively. Meanwhile, the advanced systems showed even higher sensitivity from 0.9994 to 1. The genotypes of four out of 10,000 non-relative pairs were not contradictory as parent–child pairs in the Identifiler analysis; they showed combined LRs of 566, 165, 33.6, and 21.7, which may result in false positives depending on the cutoff. When the advanced systems were applied to, all non-relative pairs shared zero alleles in at least one locus. Therefore, the employment of any advanced system or combinations thereof will be helpful for solving this type of problem if no significant LR is obtained by using Identifiler. However, it must be possible to obtain a significant result in the Identifiler analysis. 3.2. Other scenarios For the other three scenarios—full siblings vs. non-relatives, second-degree relatives vs. non-relatives, and first cousins vs. non-relatives—the distribution of the common logarithm of the combined LR is depicted in the five sets of STR loci: 24 STR loci, PowerPlex Fusion, GlobalFiler, and PowerPlex 21 (Fig. 3). Table 3 presents the sensitivity, specificity, positive predictive values, and negative predictive values. The mean value of combined LR in the relatives was raised with the increasing number of STR loci, although the mean combined LR of non-relatives declined (Supplementary Table 1). The variances of the combined LRs were raised regardless of the relatives or non-relatives. The maximum and minimum combined LRs did not depend on the number of STR loci, which was presumed to be attributable to the outliers or extreme cases of the simulation data as well as differences in STR loci included in the systems. When more STR loci were involved in, the sensitivity, specificity, positive predictive value, and negative predictive value were generally improved, because the distributions of the combined LR in the relatives and non-relatives were more clearly distinguishable. However, the relatives and non-relatives were not completely resolved as indicated in Fig. 3; in particular, especially first cousins and non-relatives showed considerably overlapping distributions with inefficient separation. 3.2.1. Full siblings The AABB cutoff LR in sibship testing is 6100 [6]. When this cutoff was adopted, the sensitivity/specificity for Identifiler, PowerPlex Fusion, and 24 STR loci were 0.8317/0.9993, 0.9474/ 0.9999, and 0.9651/0.9999, respectively. Moreover, 235 out of
Table 1 Range of combined LR (common logarithm) in parent–child. Multiplex
Maximum
Minimum
Mean ± SD
24 STR loci PowerPlex Fusion GlobalFiler PowerPlex 21 Identifiler
15.77 14.01 13.86 13.51 10.29
3.63 2.32 2.85 2.48 1.30
8.55 ± 1.53 7.38 ± 1.48 7.18 ± 1.44 7.02 ± 1.40 4.79 ± 1.20
3
the 10,000 full sibling pairs showed a combined LR <1 in the Identifiler analysis. A combined LR P 100 was achieved in 77 (32.8%) of the 235 pairs by using PowerPlex Fusion and 124 (52.8%) by using the 24 STR loci. The addition of STR loci considerably improved the combined LR for full siblings. Even with a combined LR <1 in the Identifiler analysis, a significant result can be expected by employing the advanced systems. Therefore, the results suggest these systems will be particularly advantageous in scenarios involving full siblings. 3.2.2. Second-degree relatives At an LR cutoff of 10, which is regarded as appropriate evidence indicating second-degree relative status (i.e., half-siblings) [6], the sensitivity/specificity for Identifiler, PowerPlex Fusion, and the 24 STR loci were 0.5171/0.9846, 0.6886/0.9873, and 0.7351/0.9884, respectively. In the Identifiler analysis, up to 1488 of 10,000 pairs of second-degree relatives had a combined LR <1. A combined LR P 10 was achieved in 203 (13.6%) of 1488 pairs by PowerPlex Fusion and 404 (27.2%) by the 24 STR loci. The results show that the usage of any of the advanced systems or combinations thereof will improve the analytic results to some degree, although no dramatic improvements should be expected. 3.2.3. First cousins For example, at an LR cutoff of 10, the sensitivity/specificity for Identifiler, PowerPlex Fusion, and the 24 STR loci were 0.1064/ 0.9956, 0.2143/0.9912, and 0.2633/0.9889, respectively. No marked improvements were usually expected as a result of usage of the advanced systems (data not shown). Nothnagel et al. [10] performed pairwise kinship analysis using as many as 34 autosomal STRs. They report that estimations up to second-degree relatives are practical, and state that more distant relatives must be estimated by including additional individuals. The present results also demonstrate that it is difficult to obtain significant results for first cousins even though the advanced systems are applied to. 3.3. Mutational events The addition of STR loci in parentage testing is expected to increase genetic inconsistencies because of mutations that occur during meiosis. For example, using the average mutation rate of 0.0021 of tetranucleotide STRs [15], the probability of mutation in at least one out of 24 STR loci is 0.049 compared to 0.031 with 15 STR loci; the former is about 1.6 times that of the latter. Therefore, the increase in mutational events should be kept in mind when additional STR loci are used. If genotypic inconsistencies are observed only at one locus, then the occurrence of mutation is considered, and parentage is not excluded [16]. Falsenegative results can be prevented by performing another paternal or maternal lineage testing based on Y-chromosome STRs or mitochondrial DNA. Even if two individuals are truly non-relatives, it is possible they have genotypes consistent with a parent–child relationship except for a lack of allele sharing at one locus. In this simulation study, this occurred in 37 of the 10,000 pairs in the Identifiler analysis; 28 of these pairs were explainable as an apparent one-step mutation. In contrast, this occurred in only two of 10,000 pairs when the advanced systems were used; both pairs were explainable as an apparent one-step mutation. Therefore, additional STR loci will reduce false-positive results with the accuracy of parentage testing. However, as Identifiler has usually sufficiently proved parentage in practice, the application of advanced systems to parentage testing would only confer a limited advantage.
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
4
T. Tamura et al. / Legal Medicine xxx (2015) xxx–xxx
Table 2 Sensitivity in pairwise parentage testing. Cutoff of likelihood ratio
24 STR loci
PowerPlexFusion
GlobalFiler
PowerPlex 21
Identifiler
10 100 500 1000 10,000
1.0000 1.0000 1.0000 1.0000 0.9996
1.0000 1.0000 0.9999 0.9998 0.9952
1.0000 1.0000 1.0000 0.9996 0.9927
1.0000 1.0000 0.9997 0.9994 0.9915
1.0000 0.9963 0.9743 0.9461 0.7311
Fig. 3. Distribution of combined LR (common logarithm) in pairwise kinship analysis. H1: the two individuals are relatives; H2: the two individuals are non-relatives. The simulations were performed under H1 and H2.
3.4. Linkage between loci More than 20 STR loci in a multiplex system imply the involvement of several STR pairs located on the same chromosome. In contrast to only one pair of closely located loci in Identifiler (CSF1PO–D5S818), PowerPlex Fusion includes four pairs: D2S441–TPOX, CSF1PO–D5S818, D12S391–vWA, and D21S11– Penta D; GlobalFiler has three pairs: D2S441–TPOX, CSF1PO– D5S818, and D12S391–vWA; and PowerPlex 21 has three pairs: CSF1PO–D5S818, D12S391–vWA, and D21S11–Penta D. When these advanced systems are combined together, there are five pairs: D2S441–TPOX, CSF1PO–D5S818, D6S1043–SE33, D12S391– vWA, and D21S11–Penta D. Regarding genetic distance, D2S441 and TPOX are 88.8 cM apart on chromosome 2, CSF1PO and D5S818 are 27.8 cM apart on chromosome 5, D6S1043 and SE33 are 4.4 cM apart on chromosome 6, D12S391 and vWA are 11.9 cM apart on chromosome 12, and D21S11 and Penta D are 44.7 cM apart on chromosome 21 [17]. The recombination rates of <0.5 in these locus pairs do not mean independent assortment.
When closely located loci were involved, it is necessary to consider both linkage and linkage disequilibrium. Linkage is the genetic phenomenon and describes the situation where two genes located close to each other on a chromosome are inherited together during meiosis. Linkage disequilibrium means the nonrandom association of alleles at different loci within a population. There are a couple of detailed reports concerning D12S391 and vWA [18–20]. The low-level linkage disequilibrium between the two loci for estimating the match probability for non-relatives is negligible unless a recent evolutionary event has occurred (e.g., the admixture of two substantially different populations). However, the linkage potentially affects the estimation of close relatives. If the effect of linkage in two close relatives (e.g., full-siblings, second degree-relatives) is ignored, many of LR calculation result in underestimation. The median of the ratio LR (two loci were assumed to be unlinked)/LR (these were assumed to be linked) did not exceed approximately 1.0 for above mentioned five pairs of linked loci. But, this ratio occasionally exceeded 1.0 (data not shown). This tendency has been reported in previous studies
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
LR cutoff
a
c d
PowerPlex Fusion
GlobalFiler1
PowerPlex 21
Identifiler
SPEb
PPVc
NPVd
SEN
SPE
PPV
NPV
SEN
SPE
PPV
NPV
SEN
SPE
PPV
NPV
SEN
SPE
PPV
NPV
0.9949 0.9858 0.9651 0.9387 0.9232
0.9966 0.9991 0.9999 1.0000 1.0000
0.9966 0.9991 0.9999 1.0000 1.0000
0.9949 0.9860 0.9663 0.9422 0.9287
0.9930 0.9790 0.9474 0.9104 0.8880
0.9949 0.9990 0.9999 0.9999 1.0000
0.9949 0.9990 0.9999 0.9999 1.0000
0.9930 0.9794 0.9500 0.9178 0.8993
0.9919 0.9759 0.9429 0.9013 0.8780
0.9922 0.9990 0.9999 1.0000 1.0000
0.9922 0.9990 0.9999 1.0000 1.0000
0.9919 0.9764 0.9460 0.9102 0.8913
0.9914 0.9748 0.9379 0.8961 0.8709
0.9933 0.9980 0.9996 0.9999 1.0000
0.9933 0.9980 0.9996 0.9999 1.0000
0.9914 0.9754 0.9415 0.9059 0.8857
0.9765 0.9248 0.8317 0.7319 0.6788
0.9817 0.9963 0.9993 0.9999 1.0000
0.9816 0.9960 0.9992 0.9999 1.0000
0.9766 0.9298 0.8559 0.7886 0.7569
Second-degree relatives 1 0.9157 10 0.7351 100 0.4835 500 0.3031 1000 0.2352
0.9300 0.9884 0.9992 0.9999 1.0000
0.9290 0.9845 0.9983 0.9997 1.0000
0.9169 0.7886 0.6592 0.5893 0.5666
0.9000 0.6886 0.3918 0.2144 0.1558
0.9095 0.9873 0.9992 0.9999 0.9999
0.9086 0.9819 0.9980 0.9995 0.9994
0.9009 0.7602 0.6216 0.5600 0.5422
0.9018 0.6797 0.3799 0.2073 0.1497
0.9110 0.9848 0.9992 1.0000 1.0000
0.9102 0.9781 0.9979 1.0000 1.0000
0.9027 0.7546 0.6171 0.5578 0.5405
0.8977 0.6728 0.3749 0.1926 0.1351
0.9061 0.9873 0.9988 0.9999 1.0000
0.9053 0.9815 0.9968 0.9995 1.0000
0.8986 0.7511 0.6151 0.5533 0.5362
0.8512 0.5171 0.1846 0.0639 0.0382
0.8611 0.9846 0.9994 1.0000 1.0000
0.8597 0.9711 0.9968 1.0000 1.0000
0.8527 0.6709 0.5507 0.5165 0.5097
First cousins 1 10 100 500 1000
0.7936 0.9889 0.9999 1.0000 1.0000
0.7843 0.9595 0.9976 1.0000 1.0000
0.7608 0.5731 0.5107 0.5023 0.5011
0.7332 0.2143 0.0295 0.0064 0.0034
0.7693 0.9912 0.9999 1.0000 1.0000
0.7607 0.9606 0.9966 1.0000 1.0000
0.7425 0.5578 0.5075 0.5016 0.5009
0.7310 0.2093 0.0268 0.0056 0.0023
0.7752 0.9897 1.0000 1.0000 1.0000
0.7648 0.9531 1.0000 1.0000 1.0000
0.7424 0.5559 0.5068 0.5014 0.5006
0.7268 0.1963 0.0236 0.0048 0.0021
0.7681 0.9908 1.0000 1.0000 1.0000
0.7581 0.9552 1.0000 1.0000 1.0000
0.7376 0.5521 0.5060 0.5012 0.5005
0.6854 0.1064 0.0086 0.0017 0.0007
0.7262 0.9956 1.0000 1.0000 1.0000
0.7146 0.9603 1.0000 1.0000 1.0000
0.6977 0.5270 0.5022 0.5004 0.5002
Full siblings 1 10 100 500 1000
b
24 STR loci SENa
0.7505 0.2633 0.0421 0.0093 0.0045
T. Tamura et al. / Legal Medicine xxx (2015) xxx–xxx
Sensitivity: probability of judging relatives correctly as relatives. Specificity: probability of judging non-relatives correctly as non-relatives. Positive predictive value: proportion of subjects correctly judged as relatives. Negative predictive value: proportion of subjects correctly judged as non-relatives.
5
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
Table 3 Sensitivity, specificity, positive predictive value, and negative predictive value of pairwise kinship analyses.
6
T. Tamura et al. / Legal Medicine xxx (2015) xxx–xxx
[19,21]. It is possible to judge whether the linkage affects the LR on the basis of the assumption that the genotype data of a population is in linkage equilibrium. Linkage is said to have no effects on a pedigree analysis unless (1) at least one individual in the pedigree (the central individual) is involved in at least two transmissions of genetic material, either as a parent or a child; and (2) the central individual is a double heterozygote at the loci in question [20]. The pedigree that fulfils the condition (1) corresponds to full siblings, grandparent–grandchild, uncle/aunt–nephew/niece, halfsiblings, and first cousins. Meanwhile, paternity trio and duo do not meet the condition (1) [3]. For the condition (2) in pairwise kinship analysis, it is the fundamental problem that the genotype of the central individual is hardly ever determined from the genotypes of two members in the pedigree. Therefore, the possibility of the condition (2) is not excluded. In such cases, the LR should be calculated considering linkage in addition to that ignoring linkage. For LR calculation involving a pair of linked loci, it has been suggested that either one of the two loci were used to circumvent the adverse effect of linkage [22]. But, Gill et al. [20] reported that this approach might not always conservative and recommended alternative approach using the recombination rate. A method for incorporating the recombination rate into the calculation of joint or conditional probability has been reported previously [4,23]; furthermore, Famlink freeware (http://www.famlink.se/) can calculate LRs accounting and not accounting for linkage [21]. The genetic distance >11.9 cM between D12S391 and vWA appears to have a small impact on the LR. However, the linkage between D6S1043 and SE33 (4.4 cM apart) may affect the LR much more. In addition, even though one pair of linked loci has only a small effect, two or more may give rise to a significant difference. In the advanced systems that include several pairs of linked STR loci on the same chromosome, the effects of all possible combinations on the estimation of close relatives should be comprehensively examined. 4. Conclusion This study evaluated the application of PowerPlex Fusion System, GlobalFiler Kit, and PowerPlex 21 System for pairwise kinship analysis. The authors demonstrate the simulation analysis resulting that genotyping of more than 20 autosomal STR loci improve the forensic personal identification, especially in the evaluation of full siblings and second-degree relatives, in spite of more frequent mutational events and several pairs of linked loci on the same chromosome. However, a detailed analysis should be performed to evaluate the composite impact of linkage when many more pairs of linked loci are used. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.legalmed.2015. 03.005.
References [1] Morling N, Allen RW, Carracedo A, Geada H, Guidet F, Hallenberg C, et al. Paternity Testing Commission of the International Society of Forensic Genetics: recommendations on genetic investigations in paternity cases. Forensic Sci Int 2002;129:148–57. [2] Wenk RE, Traver M, Chiafari FA. Determination of sibship in any two persons. Transfusion 1996;36:259–62. [3] Buckleton J, Triggs CM, Walsh SJ. Forensic DNA evidence interpretation. Boca Raton, FL: CRC Press; 2005. [4] Bright JA, Curran JM, Buckleton JS. Relatedness calculations for linked loci incorporating subpopulation effects. Forensic Sci Int Genet 2013;7:380–3. [5] Gjertson DW, Brenner CH, Baur MP, Carracedo A, Guidet F, Luque JA, et al. ISFG: recommendations on biostatistics in paternity testing. Forensic Sci Int Genet 2007;1:223–31. [6] American Association of Blood Banks. Annual report summary for testing in 2010 – Prepared by the Parentage Testing Standards Program Unit; 2011. Available at: http://www.aabb.org/sa/facilities/Documents/rtannrpt10.pdf [Accessed 18 September 2014]. [7] Hummel K. Biostatistical opinion of parentage. Stuttgart: Gustav Fischer Verlag; 1971. [8] 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. [9] Tamura T, Osawa M, Kimura R, Inaoka Y, Tanaka S, Satoh F, et al. Evaluation of the allele-sharing approach, known as the IBS method, in kinship analysis. J Forensic Leg Med 2013;20:112–6. [10] Nothnagel M, Schmidtke J, Krawczak M. Potentials and limits of pairwise kinship analysis using autosomal short tandem repeat loci. Int J Legal Med 2010;124:205–15. [11] Takene Y, Hashiyada M, Usui K, Hayashizaki Y, Sakai J, Funayama M. Polymorphism of 21 STRs in Japanese population. DNA Polymorphism 2012;20:260–5 (in Japanese). [12] Ozeki M, Tamaki K. Allele frequencies of 37 short tandem repeat loci in a Japanese population. Leg Med (Tokyo) 2013;15:342–6. [13] Fujii K, Iwashima Y, Kitayama T, Nakahara H, Mizuno N, Sekiguchi K. Allele frequencies for 22 autosomal short tandem repeat loci obtained by PowerPlex Fusion in a sample of 1501 individuals from the Japanese population. Leg Med (Tokyo) 2014;16:234–7. [14] R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria; 2014. Available at: http://www.R-project.org [Accessed 18 September 2014]. [15] Weber JL, Wong C. Mutation of human short tandem repeats. Hum Mol Genet 1993;2:1123–8. [16] Brenner C.H. Mutations in paternity; 2006. Available at: http://dnaview.com/mudisc.htm [Accessed 18 September 2014]. [17] Phillips C, Ballard D, Gill P, Court DS, Carracedo A, Lareu MV. Recombination landscape around forensic STRs: Accurate measurement of genetic distances between syntenic STR pairs using HapMap high density SNP data. Forensic Sci Int Genet 2012;6:354–65. [18] Phillips C, Fernandez-Formoso L, Garcia-Magariños M, Porras L, Tvedebrink T, Amigo J, et al. Analysis of global variability in 15 established and 5 new European Standard Set (ESS) STRs using the CEPH human genome diversity panel. Forensic Sci Int Genet 2011;5:155–69. [19] O’Connor KL, Tillmar AO. Effect of linkage between vWA and D12S391 in kinship analysis. Forensic Sci Int Genet 2012;6:840–4. [20] Gill P, Phillips C, McGovern C, Bright JA, Buckleton J. An evaluation of potential allelic association between the STRs vWA and D12S391: implications in criminal casework and applications to short pedigrees. Forensic Sci Int Genet 2012;6:477–86. [21] Kling D, Egeland T, Tillmar AO. FamLink–a user friendly software for linkage calculations in family genetics. Forensic Sci Int Genet 2012;6:616–20. [22] Budowle B, Ge J, Chakraborty R, Eisenberg AJ, Green R, Mulero J, et al. Population genetic analyses of the NGM STR loci. Int J Legal Med 2011;125:101–9. [23] Buckleton J, Triggs C. The effect of linkage on the calculation of DNA match probabilities for siblings and half siblings. Forensic Sci Int 2006;13:193–9.
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
Contents lists available at ScienceDirect
Legal Medicine journal homepage: www.elsevier.com/locate/legalmed
Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis Tomonori Tamura a,b, Motoki Osawa a,⇑, Eriko Ochiai a, Takanori Suzuki b, Takashi Nakamura b a b
Department of Forensic Medicine, Tokai University School of Medicine, Shimokasuya 143, Isehara, Kanagawa 259-1193, Japan Scientific Crime Laboratory, Kanagawa Prefectural Police, Yamashitacho 155, Naka-ku, Yokohama 231-0023, Japan
a r t i c l e
i n f o
Article history: Received 24 September 2014 Received in revised form 23 March 2015 Accepted 23 March 2015 Available online xxxx Keywords: Kinship analysis Short tandem repeat Likelihood ratio PowerPlex Fusion GlobalFiler PowerPlex 21
a b s t r a c t The AmpFLSTR Identifiler Kit, comprising 15 autosomal short tandem repeat (STR) loci, is commonly employed in forensic practice for calculating match probabilities and parentage testing. The conventional system exhibits insufficient estimation for kinship analysis such as sibship testing because of shortness of examined loci. This study evaluated the power of the PowerPlex Fusion System, GlobalFiler Kit, and PowerPlex 21 System, which comprise more than 20 autosomal STR loci, to estimate pairwise blood relatedness (i.e., parent–child, full siblings, second-degree relatives, and first cousins). The genotypes of all 24 STR loci in 10,000 putative pedigrees were constructed by simulation. The likelihood ratio for each locus was calculated from joint probabilities for relatives and non-relatives. The combined likelihood ratio was calculated according to the product rule. The addition of STR loci improved separation between relatives and non-relatives. However, these systems were less effectively extended to the inference for first cousins. In conclusion, these advanced systems will be useful in forensic personal identification, especially in the evaluation of full siblings and second-degree relatives. Moreover, the additional loci may give rise to two major issues of more frequent mutational events and several pairs of linked loci on the same chromosome. Ó 2015 Published by Elsevier Ireland Ltd.
1. Introduction Pedigree analysis in forensic science plays a major role in the identification of missing persons in mass disasters and accidents. In cases in which the personal remains of the deceased are unavailable, indirect identification is often performed by confirmation of any kinship with a potential family member. In addition to parentage testing, DNA analysis for other relatives such as full siblings, second-degree relatives, and first cousins is sometimes required in Japan. The clarification of the relationship between two individuals is required in the most common situation. For instance, forensic investigators are asked questions such as if the alleged father is the true father of the child or not and whether two males are brothers or not. The Paternity Testing Commission of the International Society of Forensic Genetics (ISFG) recommends the likelihood ratio (LR) approach if the weight of genetic evidence is calculated in parentage testing [1], although other parameters Abbreviations: AABB, American Association of Blood Banks; ISFG, International Society of Forensic Genetics; STR, short tandem repeat; LR, likelihood ratio. ⇑ Corresponding author. Tel.: +81 463 93 1121; fax: +81 463 92 0284. E-mail address: [email protected] (M. Osawa).
are also utilized for parentage testing (e.g., exclusion probability and probability based on Bayes’ theorem). The LR approach can be extended to other relationships such as full and half siblings [2,3]. When the genotypes used are mutually independent, the combined LR is calculated by multiplying LRs according to the product rule. If necessary, the LR and combined LR are adjusted for the effects of subpopulation and linkage [4]. Parentage is inferred according to typical cutoffs recommended by the ISFG: 100 or 1000 [5]. In sibship testing, a cutoff of 6100 is usually acceptable according to the American Association of Blood Banks (AABB) [6]. However, there is no universal cutoff among laboratories. An LR of 500 [7], which is Hummel’s paternity criterion during the blood group typing, is preferred in Japan regardless of the type of kinship. Genetic polymorphisms of short tandem repeat (STR) play a critical role in personal identification. The AmpFLSTRÒ IdentifilerÒ PCR Amplification Kit (Applied Biosystems, Foster City, CA, USA), which is composed of 15 autosomal STR loci (13 common loci of CODIS, D2S1338, and D19S433) and the Amelogenin locus, has been adopted as the nationwide standard in Japan. This system is used for the match probability calculation as well as LR calculation in duo and trio parentage testing.
http://dx.doi.org/10.1016/j.legalmed.2015.03.005 1344-6223/Ó 2015 Published by Elsevier Ireland Ltd.
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
2
T. Tamura et al. / Legal Medicine xxx (2015) xxx–xxx
However, Tamaki et al. previously found that this kit occasionally has limitations when obtaining a significant LR in pairwise sibship testing [8,9]. This limitation can be overcome by analyzing additional autosomal STR loci [10]. There are three commercially available major multiplex STR systems that comprise more than 20 autosomal STR loci: the PowerPlexÒ Fusion System (Promega Corporation, Madison, WI, USA), which comprises 22 autosomal STR loci of 13 CODIS, D1S1656, D2S441, D2S1338, D10S1248, D12S391, D19S433, D22S1045, Penta D, and Penta E; the GlobalFilerÒ PCR Amplification Kit (Applied Biosystems), which comprises 21 autosomal STR loci of 13 CODIS, D1S1656, D2S441, D2S1338, SE33, D10S1248, D12S391, D19S433, and D22S1045; and the PowerPlexÒ 21 System (Promega Corporation), which comprises 20 autosomal STR loci of 13 CODIS, D1S1656, D2S1338, D6S1043, D12S391, D19S433, Penta D, and Penta E. These three systems include five to seven loci in addition to those in the Identifiler. This study evaluated the power of the PowerPlex Fusion, GlobalFiler, and PowerPlex 21 to solve different cases of pairwise kinship analysis using simulated genotype data. In addition, the analyses were performed with the maximum set of 24 autosomal STR loci included in the three advanced systems. Finally, we discussed the problems of mutation and linkage associated with the addition of examined loci. 2. Materials and methods 2.1. Simulated STR type data for putative pedigrees For the simulation study, 10,000 datasets of five different pairwise relations of parent–child, full siblings, second-degree relatives, first cousins and non-relatives were constructed under a putative pedigree chart, representatively showed in Fig. 1. Individuals #1 and #6 represent the sets as parent–child, #5 and #6 as full siblings, #5 and #7 as second-degree relatives, #7 and #8 as first cousins, and #3 and #4 as non-relatives. Second-degree relatives are two meioses away from a particular individual in a pedigree (i.e., grandparent–grandchild, uncle/aunt–nephew/niece, and half-siblings). First, a series of genotypes in individuals #1– #4 were constructed on the basis of the allele frequencies of the 24 STR loci in the Japanese population [11–13]; then, those in individuals #5–#8 were constructed by mating of those of #1– #4. Mutational events were not considered in this process. Uniform random numbers were produced using R version 3.1.0 [14], and the construction of STR type data in the putative pedigrees were conducted in Microsoft Excel (2010).
relatives (H1) vs. non-relatives (H2), and first cousins (H1) vs. nonrelatives (H2)—were investigated by simulations. The simulation was performed with 10,000 sets of STR types where hypothesis H1 was assumed to be true, and those where hypothesis H2 was assumed to be true. The LR for each locus was calculated by dividing the joint probability assuming the two individuals were relatives by that assuming the two individuals were non-relatives on the basis of the allele frequencies in the Japanese population. For the five sets of the 24 STR loci, PowerPlex Fusion, GlobalFiler, PowerPlex 21, and Identifiler, the combined LRs were calculated by multiplying LRs based on the product rule. The linkage and subpopulation were not considered in this calculation. The combined LR was calculated using Microsoft Excel (2010). 2.3. Distribution of combined LRs, sensitivity, and specificity The distribution of the combined LR was depicted as a density curve using R version 3.1.0. In addition, sensitivity (the probability of judging relatives correctly as relatives), specificity (the probability of judging non-relatives correctly as non-relatives), positive predictive value (the proportion of subjects correctly judged as relatives), and negative predictive value (the proportion of subjects correctly judged as non-relatives) were calculated for different cutoffs of the combined LR. The power of the 24 STR loci, PowerPlex Fusion, GlobalFiler, and PowerPlex 21 to solve cases in the above mentioned scenarios was evaluated by comparing the sensitivity, specificity, positive predictive value, and negative predictive value of these advanced systems with those of Identifiler. All values were calculated using Microsoft Excel (2010). This project was approved by the Ethics Committee of the Tokai University School of Medicine. 3. Results and discussion 3.1. Parent–child scenario For the five sets of 24 STR loci, PowerPlex Fusion, GlobalFiler, PowerPlex 21, and Identifiler, the distribution of the common logarithm of the combined LR observed in the putative parent–child
2.2. Combined LR Four different case scenarios—parent–child (H1) vs. nonrelatives (H2), full siblings (H1) vs. non-relatives (H2), second-degree
#3
#8
#1
#2
#5
#6
#4
#7
Fig. 1. Putative pedigree chart for the construction of pairwise kinship.
Fig. 2. Distribution of combined LR (common logarithm) in parent–child.
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
T. Tamura et al. / Legal Medicine xxx (2015) xxx–xxx
relation is indicated in Fig. 2. The maximum, minimum, and mean ± SD of the combined LR calculated in the parent–child are summarized in Table 1 and the sensitivity is presented in Table 2. The mean, variance, and maximum of the combined LR were raised with the increasing number of STR loci. The minimum combined LR did not depend on the number of STR loci, which was presumed to be attributable to the outliers or extreme cases of the simulation data as well as differences in STR loci included in the systems. The additional STR loci generally improved sensitivity. If a standard cutoff of the combined LR of 100 or 1000 [5] was adopted, even Identifiler showed sufficiently high sensitivity up to 0.9963 and 0.9461, respectively. Meanwhile, the advanced systems showed even higher sensitivity from 0.9994 to 1. The genotypes of four out of 10,000 non-relative pairs were not contradictory as parent–child pairs in the Identifiler analysis; they showed combined LRs of 566, 165, 33.6, and 21.7, which may result in false positives depending on the cutoff. When the advanced systems were applied to, all non-relative pairs shared zero alleles in at least one locus. Therefore, the employment of any advanced system or combinations thereof will be helpful for solving this type of problem if no significant LR is obtained by using Identifiler. However, it must be possible to obtain a significant result in the Identifiler analysis. 3.2. Other scenarios For the other three scenarios—full siblings vs. non-relatives, second-degree relatives vs. non-relatives, and first cousins vs. non-relatives—the distribution of the common logarithm of the combined LR is depicted in the five sets of STR loci: 24 STR loci, PowerPlex Fusion, GlobalFiler, and PowerPlex 21 (Fig. 3). Table 3 presents the sensitivity, specificity, positive predictive values, and negative predictive values. The mean value of combined LR in the relatives was raised with the increasing number of STR loci, although the mean combined LR of non-relatives declined (Supplementary Table 1). The variances of the combined LRs were raised regardless of the relatives or non-relatives. The maximum and minimum combined LRs did not depend on the number of STR loci, which was presumed to be attributable to the outliers or extreme cases of the simulation data as well as differences in STR loci included in the systems. When more STR loci were involved in, the sensitivity, specificity, positive predictive value, and negative predictive value were generally improved, because the distributions of the combined LR in the relatives and non-relatives were more clearly distinguishable. However, the relatives and non-relatives were not completely resolved as indicated in Fig. 3; in particular, especially first cousins and non-relatives showed considerably overlapping distributions with inefficient separation. 3.2.1. Full siblings The AABB cutoff LR in sibship testing is 6100 [6]. When this cutoff was adopted, the sensitivity/specificity for Identifiler, PowerPlex Fusion, and 24 STR loci were 0.8317/0.9993, 0.9474/ 0.9999, and 0.9651/0.9999, respectively. Moreover, 235 out of
Table 1 Range of combined LR (common logarithm) in parent–child. Multiplex
Maximum
Minimum
Mean ± SD
24 STR loci PowerPlex Fusion GlobalFiler PowerPlex 21 Identifiler
15.77 14.01 13.86 13.51 10.29
3.63 2.32 2.85 2.48 1.30
8.55 ± 1.53 7.38 ± 1.48 7.18 ± 1.44 7.02 ± 1.40 4.79 ± 1.20
3
the 10,000 full sibling pairs showed a combined LR <1 in the Identifiler analysis. A combined LR P 100 was achieved in 77 (32.8%) of the 235 pairs by using PowerPlex Fusion and 124 (52.8%) by using the 24 STR loci. The addition of STR loci considerably improved the combined LR for full siblings. Even with a combined LR <1 in the Identifiler analysis, a significant result can be expected by employing the advanced systems. Therefore, the results suggest these systems will be particularly advantageous in scenarios involving full siblings. 3.2.2. Second-degree relatives At an LR cutoff of 10, which is regarded as appropriate evidence indicating second-degree relative status (i.e., half-siblings) [6], the sensitivity/specificity for Identifiler, PowerPlex Fusion, and the 24 STR loci were 0.5171/0.9846, 0.6886/0.9873, and 0.7351/0.9884, respectively. In the Identifiler analysis, up to 1488 of 10,000 pairs of second-degree relatives had a combined LR <1. A combined LR P 10 was achieved in 203 (13.6%) of 1488 pairs by PowerPlex Fusion and 404 (27.2%) by the 24 STR loci. The results show that the usage of any of the advanced systems or combinations thereof will improve the analytic results to some degree, although no dramatic improvements should be expected. 3.2.3. First cousins For example, at an LR cutoff of 10, the sensitivity/specificity for Identifiler, PowerPlex Fusion, and the 24 STR loci were 0.1064/ 0.9956, 0.2143/0.9912, and 0.2633/0.9889, respectively. No marked improvements were usually expected as a result of usage of the advanced systems (data not shown). Nothnagel et al. [10] performed pairwise kinship analysis using as many as 34 autosomal STRs. They report that estimations up to second-degree relatives are practical, and state that more distant relatives must be estimated by including additional individuals. The present results also demonstrate that it is difficult to obtain significant results for first cousins even though the advanced systems are applied to. 3.3. Mutational events The addition of STR loci in parentage testing is expected to increase genetic inconsistencies because of mutations that occur during meiosis. For example, using the average mutation rate of 0.0021 of tetranucleotide STRs [15], the probability of mutation in at least one out of 24 STR loci is 0.049 compared to 0.031 with 15 STR loci; the former is about 1.6 times that of the latter. Therefore, the increase in mutational events should be kept in mind when additional STR loci are used. If genotypic inconsistencies are observed only at one locus, then the occurrence of mutation is considered, and parentage is not excluded [16]. Falsenegative results can be prevented by performing another paternal or maternal lineage testing based on Y-chromosome STRs or mitochondrial DNA. Even if two individuals are truly non-relatives, it is possible they have genotypes consistent with a parent–child relationship except for a lack of allele sharing at one locus. In this simulation study, this occurred in 37 of the 10,000 pairs in the Identifiler analysis; 28 of these pairs were explainable as an apparent one-step mutation. In contrast, this occurred in only two of 10,000 pairs when the advanced systems were used; both pairs were explainable as an apparent one-step mutation. Therefore, additional STR loci will reduce false-positive results with the accuracy of parentage testing. However, as Identifiler has usually sufficiently proved parentage in practice, the application of advanced systems to parentage testing would only confer a limited advantage.
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
4
T. Tamura et al. / Legal Medicine xxx (2015) xxx–xxx
Table 2 Sensitivity in pairwise parentage testing. Cutoff of likelihood ratio
24 STR loci
PowerPlexFusion
GlobalFiler
PowerPlex 21
Identifiler
10 100 500 1000 10,000
1.0000 1.0000 1.0000 1.0000 0.9996
1.0000 1.0000 0.9999 0.9998 0.9952
1.0000 1.0000 1.0000 0.9996 0.9927
1.0000 1.0000 0.9997 0.9994 0.9915
1.0000 0.9963 0.9743 0.9461 0.7311
Fig. 3. Distribution of combined LR (common logarithm) in pairwise kinship analysis. H1: the two individuals are relatives; H2: the two individuals are non-relatives. The simulations were performed under H1 and H2.
3.4. Linkage between loci More than 20 STR loci in a multiplex system imply the involvement of several STR pairs located on the same chromosome. In contrast to only one pair of closely located loci in Identifiler (CSF1PO–D5S818), PowerPlex Fusion includes four pairs: D2S441–TPOX, CSF1PO–D5S818, D12S391–vWA, and D21S11– Penta D; GlobalFiler has three pairs: D2S441–TPOX, CSF1PO– D5S818, and D12S391–vWA; and PowerPlex 21 has three pairs: CSF1PO–D5S818, D12S391–vWA, and D21S11–Penta D. When these advanced systems are combined together, there are five pairs: D2S441–TPOX, CSF1PO–D5S818, D6S1043–SE33, D12S391– vWA, and D21S11–Penta D. Regarding genetic distance, D2S441 and TPOX are 88.8 cM apart on chromosome 2, CSF1PO and D5S818 are 27.8 cM apart on chromosome 5, D6S1043 and SE33 are 4.4 cM apart on chromosome 6, D12S391 and vWA are 11.9 cM apart on chromosome 12, and D21S11 and Penta D are 44.7 cM apart on chromosome 21 [17]. The recombination rates of <0.5 in these locus pairs do not mean independent assortment.
When closely located loci were involved, it is necessary to consider both linkage and linkage disequilibrium. Linkage is the genetic phenomenon and describes the situation where two genes located close to each other on a chromosome are inherited together during meiosis. Linkage disequilibrium means the nonrandom association of alleles at different loci within a population. There are a couple of detailed reports concerning D12S391 and vWA [18–20]. The low-level linkage disequilibrium between the two loci for estimating the match probability for non-relatives is negligible unless a recent evolutionary event has occurred (e.g., the admixture of two substantially different populations). However, the linkage potentially affects the estimation of close relatives. If the effect of linkage in two close relatives (e.g., full-siblings, second degree-relatives) is ignored, many of LR calculation result in underestimation. The median of the ratio LR (two loci were assumed to be unlinked)/LR (these were assumed to be linked) did not exceed approximately 1.0 for above mentioned five pairs of linked loci. But, this ratio occasionally exceeded 1.0 (data not shown). This tendency has been reported in previous studies
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
LR cutoff
a
c d
PowerPlex Fusion
GlobalFiler1
PowerPlex 21
Identifiler
SPEb
PPVc
NPVd
SEN
SPE
PPV
NPV
SEN
SPE
PPV
NPV
SEN
SPE
PPV
NPV
SEN
SPE
PPV
NPV
0.9949 0.9858 0.9651 0.9387 0.9232
0.9966 0.9991 0.9999 1.0000 1.0000
0.9966 0.9991 0.9999 1.0000 1.0000
0.9949 0.9860 0.9663 0.9422 0.9287
0.9930 0.9790 0.9474 0.9104 0.8880
0.9949 0.9990 0.9999 0.9999 1.0000
0.9949 0.9990 0.9999 0.9999 1.0000
0.9930 0.9794 0.9500 0.9178 0.8993
0.9919 0.9759 0.9429 0.9013 0.8780
0.9922 0.9990 0.9999 1.0000 1.0000
0.9922 0.9990 0.9999 1.0000 1.0000
0.9919 0.9764 0.9460 0.9102 0.8913
0.9914 0.9748 0.9379 0.8961 0.8709
0.9933 0.9980 0.9996 0.9999 1.0000
0.9933 0.9980 0.9996 0.9999 1.0000
0.9914 0.9754 0.9415 0.9059 0.8857
0.9765 0.9248 0.8317 0.7319 0.6788
0.9817 0.9963 0.9993 0.9999 1.0000
0.9816 0.9960 0.9992 0.9999 1.0000
0.9766 0.9298 0.8559 0.7886 0.7569
Second-degree relatives 1 0.9157 10 0.7351 100 0.4835 500 0.3031 1000 0.2352
0.9300 0.9884 0.9992 0.9999 1.0000
0.9290 0.9845 0.9983 0.9997 1.0000
0.9169 0.7886 0.6592 0.5893 0.5666
0.9000 0.6886 0.3918 0.2144 0.1558
0.9095 0.9873 0.9992 0.9999 0.9999
0.9086 0.9819 0.9980 0.9995 0.9994
0.9009 0.7602 0.6216 0.5600 0.5422
0.9018 0.6797 0.3799 0.2073 0.1497
0.9110 0.9848 0.9992 1.0000 1.0000
0.9102 0.9781 0.9979 1.0000 1.0000
0.9027 0.7546 0.6171 0.5578 0.5405
0.8977 0.6728 0.3749 0.1926 0.1351
0.9061 0.9873 0.9988 0.9999 1.0000
0.9053 0.9815 0.9968 0.9995 1.0000
0.8986 0.7511 0.6151 0.5533 0.5362
0.8512 0.5171 0.1846 0.0639 0.0382
0.8611 0.9846 0.9994 1.0000 1.0000
0.8597 0.9711 0.9968 1.0000 1.0000
0.8527 0.6709 0.5507 0.5165 0.5097
First cousins 1 10 100 500 1000
0.7936 0.9889 0.9999 1.0000 1.0000
0.7843 0.9595 0.9976 1.0000 1.0000
0.7608 0.5731 0.5107 0.5023 0.5011
0.7332 0.2143 0.0295 0.0064 0.0034
0.7693 0.9912 0.9999 1.0000 1.0000
0.7607 0.9606 0.9966 1.0000 1.0000
0.7425 0.5578 0.5075 0.5016 0.5009
0.7310 0.2093 0.0268 0.0056 0.0023
0.7752 0.9897 1.0000 1.0000 1.0000
0.7648 0.9531 1.0000 1.0000 1.0000
0.7424 0.5559 0.5068 0.5014 0.5006
0.7268 0.1963 0.0236 0.0048 0.0021
0.7681 0.9908 1.0000 1.0000 1.0000
0.7581 0.9552 1.0000 1.0000 1.0000
0.7376 0.5521 0.5060 0.5012 0.5005
0.6854 0.1064 0.0086 0.0017 0.0007
0.7262 0.9956 1.0000 1.0000 1.0000
0.7146 0.9603 1.0000 1.0000 1.0000
0.6977 0.5270 0.5022 0.5004 0.5002
Full siblings 1 10 100 500 1000
b
24 STR loci SENa
0.7505 0.2633 0.0421 0.0093 0.0045
T. Tamura et al. / Legal Medicine xxx (2015) xxx–xxx
Sensitivity: probability of judging relatives correctly as relatives. Specificity: probability of judging non-relatives correctly as non-relatives. Positive predictive value: proportion of subjects correctly judged as relatives. Negative predictive value: proportion of subjects correctly judged as non-relatives.
5
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005
Table 3 Sensitivity, specificity, positive predictive value, and negative predictive value of pairwise kinship analyses.
6
T. Tamura et al. / Legal Medicine xxx (2015) xxx–xxx
[19,21]. It is possible to judge whether the linkage affects the LR on the basis of the assumption that the genotype data of a population is in linkage equilibrium. Linkage is said to have no effects on a pedigree analysis unless (1) at least one individual in the pedigree (the central individual) is involved in at least two transmissions of genetic material, either as a parent or a child; and (2) the central individual is a double heterozygote at the loci in question [20]. The pedigree that fulfils the condition (1) corresponds to full siblings, grandparent–grandchild, uncle/aunt–nephew/niece, halfsiblings, and first cousins. Meanwhile, paternity trio and duo do not meet the condition (1) [3]. For the condition (2) in pairwise kinship analysis, it is the fundamental problem that the genotype of the central individual is hardly ever determined from the genotypes of two members in the pedigree. Therefore, the possibility of the condition (2) is not excluded. In such cases, the LR should be calculated considering linkage in addition to that ignoring linkage. For LR calculation involving a pair of linked loci, it has been suggested that either one of the two loci were used to circumvent the adverse effect of linkage [22]. But, Gill et al. [20] reported that this approach might not always conservative and recommended alternative approach using the recombination rate. A method for incorporating the recombination rate into the calculation of joint or conditional probability has been reported previously [4,23]; furthermore, Famlink freeware (http://www.famlink.se/) can calculate LRs accounting and not accounting for linkage [21]. The genetic distance >11.9 cM between D12S391 and vWA appears to have a small impact on the LR. However, the linkage between D6S1043 and SE33 (4.4 cM apart) may affect the LR much more. In addition, even though one pair of linked loci has only a small effect, two or more may give rise to a significant difference. In the advanced systems that include several pairs of linked STR loci on the same chromosome, the effects of all possible combinations on the estimation of close relatives should be comprehensively examined. 4. Conclusion This study evaluated the application of PowerPlex Fusion System, GlobalFiler Kit, and PowerPlex 21 System for pairwise kinship analysis. The authors demonstrate the simulation analysis resulting that genotyping of more than 20 autosomal STR loci improve the forensic personal identification, especially in the evaluation of full siblings and second-degree relatives, in spite of more frequent mutational events and several pairs of linked loci on the same chromosome. However, a detailed analysis should be performed to evaluate the composite impact of linkage when many more pairs of linked loci are used. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.legalmed.2015. 03.005.
References [1] Morling N, Allen RW, Carracedo A, Geada H, Guidet F, Hallenberg C, et al. Paternity Testing Commission of the International Society of Forensic Genetics: recommendations on genetic investigations in paternity cases. Forensic Sci Int 2002;129:148–57. [2] Wenk RE, Traver M, Chiafari FA. Determination of sibship in any two persons. Transfusion 1996;36:259–62. [3] Buckleton J, Triggs CM, Walsh SJ. Forensic DNA evidence interpretation. Boca Raton, FL: CRC Press; 2005. [4] Bright JA, Curran JM, Buckleton JS. Relatedness calculations for linked loci incorporating subpopulation effects. Forensic Sci Int Genet 2013;7:380–3. [5] Gjertson DW, Brenner CH, Baur MP, Carracedo A, Guidet F, Luque JA, et al. ISFG: recommendations on biostatistics in paternity testing. Forensic Sci Int Genet 2007;1:223–31. [6] American Association of Blood Banks. Annual report summary for testing in 2010 – Prepared by the Parentage Testing Standards Program Unit; 2011. Available at: http://www.aabb.org/sa/facilities/Documents/rtannrpt10.pdf [Accessed 18 September 2014]. [7] Hummel K. Biostatistical opinion of parentage. Stuttgart: Gustav Fischer Verlag; 1971. [8] 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. [9] Tamura T, Osawa M, Kimura R, Inaoka Y, Tanaka S, Satoh F, et al. Evaluation of the allele-sharing approach, known as the IBS method, in kinship analysis. J Forensic Leg Med 2013;20:112–6. [10] Nothnagel M, Schmidtke J, Krawczak M. Potentials and limits of pairwise kinship analysis using autosomal short tandem repeat loci. Int J Legal Med 2010;124:205–15. [11] Takene Y, Hashiyada M, Usui K, Hayashizaki Y, Sakai J, Funayama M. Polymorphism of 21 STRs in Japanese population. DNA Polymorphism 2012;20:260–5 (in Japanese). [12] Ozeki M, Tamaki K. Allele frequencies of 37 short tandem repeat loci in a Japanese population. Leg Med (Tokyo) 2013;15:342–6. [13] Fujii K, Iwashima Y, Kitayama T, Nakahara H, Mizuno N, Sekiguchi K. Allele frequencies for 22 autosomal short tandem repeat loci obtained by PowerPlex Fusion in a sample of 1501 individuals from the Japanese population. Leg Med (Tokyo) 2014;16:234–7. [14] R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria; 2014. Available at: http://www.R-project.org [Accessed 18 September 2014]. [15] Weber JL, Wong C. Mutation of human short tandem repeats. Hum Mol Genet 1993;2:1123–8. [16] Brenner C.H. Mutations in paternity; 2006. Available at: http://dnaview.com/mudisc.htm [Accessed 18 September 2014]. [17] Phillips C, Ballard D, Gill P, Court DS, Carracedo A, Lareu MV. Recombination landscape around forensic STRs: Accurate measurement of genetic distances between syntenic STR pairs using HapMap high density SNP data. Forensic Sci Int Genet 2012;6:354–65. [18] Phillips C, Fernandez-Formoso L, Garcia-Magariños M, Porras L, Tvedebrink T, Amigo J, et al. Analysis of global variability in 15 established and 5 new European Standard Set (ESS) STRs using the CEPH human genome diversity panel. Forensic Sci Int Genet 2011;5:155–69. [19] O’Connor KL, Tillmar AO. Effect of linkage between vWA and D12S391 in kinship analysis. Forensic Sci Int Genet 2012;6:840–4. [20] Gill P, Phillips C, McGovern C, Bright JA, Buckleton J. An evaluation of potential allelic association between the STRs vWA and D12S391: implications in criminal casework and applications to short pedigrees. Forensic Sci Int Genet 2012;6:477–86. [21] Kling D, Egeland T, Tillmar AO. FamLink–a user friendly software for linkage calculations in family genetics. Forensic Sci Int Genet 2012;6:616–20. [22] Budowle B, Ge J, Chakraborty R, Eisenberg AJ, Green R, Mulero J, et al. Population genetic analyses of the NGM STR loci. Int J Legal Med 2011;125:101–9. [23] Buckleton J, Triggs C. The effect of linkage on the calculation of DNA match probabilities for siblings and half siblings. Forensic Sci Int 2006;13:193–9.
Please cite this article in press as: Tamura T et al. Evaluation of advanced multiplex short tandem repeat systems in pairwise kinship analysis. Leg Med (2015), http://dx.doi.org/10.1016/j.legalmed.2015.03.005