COMPARISON OF TWO DIFFERENT MULTIPLEX SYSTEMS IN CALCULATING KINSHIP AMONG CLOSE RELATIVES

This study compares the results obtained using two multiplex systems, PowerPlex ® 16 System and PowerPlex ® Fusion System, to evaluate the probability of a specific kinship relationship between the offspring of three pairs of identical twins, such as full kinship (siblings), first-degree relatives (first cousins) and half-siblings. Genomic DNA was isolated and amplified from buccal swab and selected short tandem repeat (STR) markers were detected. Electropherograms were generated and analyzed for all persons, using two multiplex systems. Paternity testing for every nine offspring of six examined couples was performed and in all cases the probability that the alleged father is the true father, was over 99.9999%. Kinship analyses were performed setting up two different hypotheses and calculating the likelihood ratio (LR) and kinship probability. Determining the degree of kinship between persons who were full siblings, likelihood ratio showed the highest values contrary to other two types of kinship. Kinship analyses between first cousins showed a higher probability that the examined persons are half-siblings, rather than they are first cousins. In most cases, the introduction of additional seven loci included in PowerPlex ® Fusion System increased the values of average likelihood ratios. It is recommendable to use over 20 STR loci in complex kinship analyses.


Introduction
Personal identification or complex kinship analyses represents the most challenging tasks for forensic investigators. Determining sibship or half sibship among close relatives is a frequent issue in kinship cases. These analyses can be carried out by testing a set of STR (Short Tandem Repeat) loci. STR allelic variants vary from person to person, so it is not unusual that two persons share the same alleles at STR loci, or even to match at two or three STR loci. However, the minimum theoretical probability that two individuals share identical alleles at all 15 STR loci included in the commercial multiplex STR system PowerPlex ® 16 for the Caucasian population is 1/1.83 x 10 17 (Marjanovic and Primorac, 2013).
STR markers became very popular in forensic practice because they are based on PCR technology, and they can be used when it comes to a small amount, or degraded DNA. STR markers are suitable for multiplex amplification and include sensitive fluorescent detection, which allows researchers to quickly collect the data based on the markers. STR markers have a great power of discrimination between persons who are not related, but also between closely related individuals (Butler, 2015).
A large number of multiplex STR systems have been developed and they are used in forensic practice for many purposes, like confirming or excluding paternity, but also, for other types of relationships between individuals. For example, Thomson et al. (2001) analyzed the kinship relations using multiplex STR loci and Gaytmenn et al. (2002) studied the sensitivity and specificity of kinship analysis. Reid et al. (2004) compared the probability of sibship for pairs of full siblings versus unrelated individuals, using Identifiler multiplex STR kit (15 STR loci plus amelogenin). In this study, LR values for known full siblings ranged from 4.6 to over 1 billion and for unrelated individuals from 0.000000045 to 0.12. All full siblings had an LR > 1 and all nonsiblings had an LR < 1.
Appropriate application of statistical model plays an important role in the determination of kinship between individuals. To calculate the kinship between two individuals, it is necessary to define the relatedness coefficients for each type of relations. The relatedness coefficients are commonly referred as k 0 , 2k 1 , k 2 (for two observed persons, X and Y): k 0 = P (none of allele X is identical by descent to Y alleles); 2k 1 = P (one allele X is identical by descent to one of the Y allele, but the second one is not); k 2 = P (both X alleles are identical by descent to Y alleles), where P is the probability for each individually analyzed locus (Fung and Hu, 2008). The probability of kinship is calculated based on these coefficients.
The aim of this study was to compare the performance of two multiplex systems, PowerPlex ® 16 and PowerPlex ® Fusion, in the process of statistical determination of kinship between the descendants of three pairs of identical twins. The specificity of this study is that the descendants of identical twins are first cousins, but regarding their DNA profiles, they are half siblings. The main objective was to investigate whether larger number of loci contained within PowerPlex ® Fusion System, increases the possibility of discrimination between three types of kinship: full sibship, first cousins and half sibship.

Materials and methods
Samples were collected from 21 individuals, of which three pairs of identical twins with their spouses and with a total of nine offspring (Fig.  1). The first twin pair is labeled as D and N, their spouses are A and T, and their children are DA1, DA2, NT1 and NT2. The second twin pair is labeled as M and B, their spouses as V and T1, and children as MV, BT1 and BT2. The third twin pair is labeled as AD and AD1, their spouses as SA and ME and their children as ASA and AMS. All persons investigated gave informed consent.  (Miller, Dykes and Polesky, 1987) and selected STR markers were detected using ABI PRISM 310 Genetic Analyzer and two multiplex systems.
Data analysis was performed using GeneMapper ® ID software, ver. 3.2. Paternity testing for all the descendants of six examined couples was performed. Kinship analyses were performed by setting up two different hypotheses and calculating the LR. Calculations were performed using EasyDNA_2Persons software, suggested by Fung and Hu (2008). Before starting the software analysis, it was necessary to make one input file, which contains a list of all 15 loci, allelic variants and their frequencies for PowerPlex ® 16 System, or 22 for PowerPlex ® Fusion System. This software uses relatedness coefficients to describe the type of relationship between two persons under the hypothesis (Tab. 1). Table 1. Relatedness coefficients (k0, 2k1, k2) for some common relationships between two persons (Fung and Hu, 2008) Hypothesis testing and calculation of LR was performed for each locus individually, using the formulas for the calculation of probability, as shown in Table 2.  As an input file for EasyDNA_2Persons software, we used the list of loci and allelic variants for the Bosnian-Herzegovinian population (Marjanovic et al., 2006). Profiles, obtained using PowerPlex ® Fusion System, include seven additional loci, whose allelic frequencies were taken from the data for Croatian population (Curic, Gasic, Pluzaric and Smiljcic, 2012).

Results and discussion
Electropherograms were generated and analyzed using two multiplex systems, where both multiplex systems gave satisfactory results for all individuals. DNA samples were profiled using PowerPlex ® 16 System and PowerPlex ® Fusion System. Samples were analyzed using the ABI PRISM 310 Genetic Analyzer and the profiles were obtained using GeneMapper ® ID software, ver. 3.2.
In all the cases of paternity testing using the PowerPlex ® 16 System, probability that the alleged father is the true father, was higher than 99.9999%. There were some variations in the values of CPI in some samples, which reflected on the probability of paternity. The chart shows that the lowest probability of paternity had parental couple labeled as M and V, for the child labeled as sample MV (99.99998%) and a parental couple labeled as B and T1, for the child labeled as sample BT1 (99.999975%) (Fig. 2).
Other samples had similar values, except for the pair of samples AD and AD1, where in both cases of paternity testing were obtained much higher probabilities, compared to the other tested samples. High CPI values resulted from rare allelic variants present at several loci. In this case, Penta E locus was especially interesting, since twin pairs with the highest probability of paternity, had allelic variant 20. According to Marjanovic et al. (2006), this allelic variant was not observed in a sample of 100 individuals in the Bosnian-Herzegovinian population, so allelic frequency of variant 20 is considered as 0.005. Another allelic variant that was not observed by Marjanovic et al. (2006) is 14.3 at the D1S1656 locus, which was found in the sample T.
Having confirmed the paternity for every descendant of all twin pairs, calculation of the degree of kinship could be performed. The examined types of kinship were: full sibship, first-degree relatives and half sibship. It was Figure 2. Paternity testing for all samples using the PowerPlex ® 16 System necessary to set up two hypotheses (Hp and Hd) and to calculate LR or probability factor. If the value of LR is greater than 1, it supports the probability of Hp hypothesis, and if the value is less than 1, it supports probability of Hd hypothesis.
Firstly, we tested the following two hypotheses: Hp -The relationship between the examined persons is full kinship (brothers and sisters) and Hd -The relationship between the examined persons is first-degree relatives (first cousins). The obtained average LRs for both multiplex systems are presented in Table 3.  Yuan et al. (2017) as loci with higher discrimination power. STRs with higher discrimination power values should be included in analyses when additional autosomal markers are required for full sibship identification. Next, we tested the following two hypotheses: Hp -The relationship between the examined persons is full kinship (brothers and sisters) and Hd -The relationship between the examined samples is half siblings (half-brothers and half-sisters). Results are presented in Table  4. Regarding the probabilities of full sibship All cases of full sibship were again confirmed, in both multiplex systems. It is also noticable that in all analyses, additional loci in the PowerPlex ® Fusion System gave higher support to supposed relations, ie. higher values for full sibship in fully related individuals and lower values of full sibship in half related individuals.
Finally, two hypotheses were tested: Hp -The relationship between the examined samples is half siblings (half-brothers and half-sisters) and Hd -The relationship between the examined persons is first-degree relatives (first cousins). The fully related individuals were excluded from this analysis (Tab. 5   The introduction of additional seven loci used in the PowerPlex ® Fusion System increased the values of LR in 5 out of 7 analyzed pairs. Carboni et al. (2014) reported a case study of two women who were cousins and they suspected to be daughters of the same father. Their mothers were sisters and were not available for kinship testing. The analysis with a commercial kit showed a probability value of about 15, to support the hypothesis that they were half-sisters. Including the additional 26Plex markers, the total of 41 loci, probability value increased to 435, supporting the hypothesis that the subjects were half-sisters.
Our results also suggest the use of larger number of loci when distinguishing between full and half sibship.

Conclusions
In this study, we compared two multiplex systems, PowerPlex ® 16 System and PowerPlex ® Fusion System, regarding the determination of kinship between the descendants of identical twins. Likelihood ratios obtained by calculating the degree of kinship between first cousins provided greater support to the hypothesis that they are half siblings. In all cases of full siblings, likelihood ratio the most strongly supported the hypothesis of full sibship.
Additional seven loci in PowerPlex ® Fusion System gave more accurate results, so it is recommendable to use 20 or more STR loci in analysis of different types of kinships to avoid ambiguous results and to facilitate interpretation of obtained results.