Forensic Science International: Genetics 4 (2009) 34–42

Validation of a single nucleotide polymorphism (SNP) typing assay with 49 SNPsfor forensic genetic testing in a laboratory accredited according to the ISO 17025standard

Claus Børsting *, Eszter Rockenbauer, Niels Morling

Section of Forensic Genetics, Department of Forensic Medicine, Faculty of Health Sciences, University of Copenhagen, 11 Frederik V’s Vej, DK-2100 Copenhagen, Denmark

A R T I C L E I N F O

Article history:

Received 23 February 2009

Received in revised form 3 April 2009

Accepted 9 April 2009

Keywords:

ISO 17025 standard

Accreditation

Single nucleotide polymorphism

Multiplex

Forensic genetics

A B S T R A C T

A multiplex assay with 49 autosomal single nucleotide polymorphisms (SNPs) developed for human

identification was validated for forensic genetic casework and accredited according to the ISO 17025

standard. The multiplex assay was based on the SNPforID 52plex SNP assay [J.J. Sanchez, C. Phillips, C.

Børsting, K. Balogh, M. Bogus, M. Fondevila, C.D. Harrison, E. Musgrave-Brown, A. Salas, D. Syndercombe-

Court, P.M. Schneider, A. Carracedo, N. Morling, A multiplex assay with 52 single nucleotide

polymorphisms for human identification, Electrophoresis 27 (2006) 1713–1724], where 52 fragments

were amplified in one PCR reaction. The SNPs were analysed by single base extension (SBE) and capillary

electrophoresis. Twenty-three of the original SBE primers were altered to improve the overall robustness

of the assay and to simplify the analysis of the SBE results. A total of 216 samples from 50 paternity cases

and 33 twin cases were typed at least twice for the 49 SNPs. All electropherograms were analysed

independently by two expert analysts prior to approval. Based on these results, detailed guidelines for

analysis of the SBE products were developed. With these guidelines, the peak height ratio of a

heterozygous allele call or the signal to noise ratio of a homozygous allele call is compared with

previously obtained ratios. A laboratory protocol for analysis of SBE products was developed where allele

calls with unusual ratios were highlighted to facilitate the analysis of difficult allele calls. The guidelines

for allele calling proved to be highly efficient for the detection of DNA mixtures and contaminated DNA

preparations. DNA from two individuals was mixed in seven different ratios ranging from 1:1 to 1:10; all

mixtures were easily identified as mixtures.

� 2009 Elsevier Ireland Ltd. All rights reserved.

Contents lists available at ScienceDirect

Forensic Science International: Genetics

journa l homepage: www.e lsev ier .com/ locate / fs ig

1. Introduction

Quality assurance and standardisation of methods have beenimportant focus areas for the forensic community since DNAinvestigations were introduced in forensic genetic casework in thesecond half of the 1980s [1]. The forensic community has issuedrecommendations on all steps of the forensic DNA investigation,from selection of loci to the calculation of the statistical weight ofthe evidence (reviewed in Ref. [2]). Furthermore, the forensiccommunity recommended that laboratories that perform forensicDNA investigations, shall be accredited according to the interna-tional laboratory standard ISO 17025 for testing and calibrationlaboratories [3–5]. This has increased the confidence in forensicDNA investigations both in the general public and within thejustice system. Nevertheless, criticism is raised occasionally whenerroneous DNA investigations are revealed or new ideas are

* Corresponding author. Tel.: +45 35 32 61 10; fax: +45 35 32 61 20.

E-mail address: claus.boersting@forensic.ku.dk (C. Børsting).

1872-4973/$ – see front matter � 2009 Elsevier Ireland Ltd. All rights reserved.

doi:10.1016/j.fsigen.2009.04.004

employed [6–8]. This emphasizes the need for continuous focus onquality assurance by the forensic community, especially when newmethods are introduced.

In recent years, the use of single nucleotide polymorphisms(SNPs) in forensic genetic casework has been widely discussed [9–11] and several autosomal SNP marker sets for the identification ofhumans were developed [12–18]. The largest of these marker setswas the 52plex SNP assay developed by the SNPforID consortium[16]. All 52 SNP loci were amplified in one PCR reaction and theSNPs were analysed by single base extension (SBE) [16]. Theperformance of the 52plex SNP assay was recently compared to theuse of STRs and VNTRs (variable number of tandem repeats) inpaternity casework [19]. The typical paternity indices (PIs)obtained based on the 52 SNPs were 5–50 times lower than thetypical PIs based on 15 STRs or 7 VNTRs. However, the mutationrate of SNPs is expected to be approximately 100,000 times lowerthan the mutation rate of tandem repeats, and this makes theSNPforID 52plex SNP assay a valuable investigative tool in forensiccasework where potential relatives are involved. Especially incases where various family scenarios are possible or in cases where

C. Børsting et al. / Forensic Science International: Genetics 4 (2009) 34–42 35

a possible mutation has been detected in one of the tandem repeatloci, and supplementary investigations are required. Furthermore,the SNPforID 52plex SNP assay may be highly useful in disastervictim investigations, because the assay can generate completeprofiles from highly degraded DNA, where the commercial STR kitsfail [16].

The SNPforID multiplex was previously tested in a number ofinter-laboratory exercises within the SNPforID consortium [20]and the European DNA Profiling Group [21]. Generally, theparticipating laboratories performed well and it seemed relativelyeasy to introduce the SBE method in forensic genetic laboratories.However, some discrepancies were observed between thelaboratories and often the discrepancies were caused by simplemisinterpretations of the electropherograms. In the SBE reaction,the SBE primers are extended with a labelled ddNTP complimen-tary to the nucleotide in the SNP position. For heterozygousindividuals, the SBE primers are labelled with two different dyesand consequently, the extended SBE primers are detected atdifferent wavelengths and appear in different dye windows of theelectropherogram. Furthermore, the strengths of the fluorophoreemissions from the four dyes used in the SNaPshot kit areunbalanced, and the peak height of one allele may be up to 6 timeshigher than the peak height of the other allele in the same locus.This is inconvenient for the analyst, especially when many SNPs areanalysed in the same electropherogram, and it increases the risk offalse interpretation of the results. Previously, this problem wasaddressed by normalising the peak height values according to theapproximate strength of the fluorescent dyes in the SNaPshot kit,and by using general principles for allele calling based on thenormalised peak heights [16,19,21]. In laboratories, where SBE wasperformed routinely, this was sufficient to obtain reproducibleresults. However, inexperienced analysts sometimes made erro-neous allele calls if the signals from the SBE products were weak orthe noise was high in a certain area of the electropherogram. One ofthe important lessons learned from the inter-laboratory exerciseswas that analysis of the SBE results required extensive training andthat it was necessary to refine the principles used for allele callingin order to obtain fully concordant results in different laboratories.

When it was decided to validate the SNPforID multiplex forforensic genetic testing and apply for ISO 17025 accreditation, themain focus of the validation was to address the problems identifiedin the inter-laboratory exercises and develop specific guidelines forallele calling that would fulfil the requirements of objective evidencestated in the management systems requirements of ISO/IEC 17025.

The work presented here is the essential part of the validationreport submitted to the Danish national body for accreditation(http://webtool.danak.dk/).

2. Materials and methods

2.1. Samples

A total of 50 paternity cases from 2005 or 2006, and 33 twincases, in which the zygosity should be determined, were selected.The selection criterion was that the samples were collected aseither buccal swabs or blood on FTA1 cards (Whatman). The workwas approved by the Danish ethical committee (KF-01-037/03).

2.2. Sample preparation

One 1.2 mm disk was punched out of the FTA card directly into amicrotitre plate (Eppendorf) containing 10 ml of ddH2O using aBSD600-duet puncher (BSD Robotics). After each sample-punch, acleaning-punch was performed. The FTA punches were washed bya THEONYX liquid handler (MWG) using the following protocol:150 ml of ddH2O was added to each well and was mixed by

aspirating and dispensing 5 times. The H2O was left in the wells fora minimum of 10 min and mixed by aspirating and dispensing 5times followed by removal of approximately 130 ml liquid. Again,150 ml of ddH2O was added to each well and mixed by aspiratingand dispensing 5 times. The H2O was left in the wells for aminimum of 10 min, and mixed by aspirating and dispensing 5times followed by removal of as much of the remaining liquid aspossible. All punches were dried at 60 8C for 45 min [22].

2.3. STR analysis

A total of 15 autosomal STRs (CSF1PO, D13S317, D16S539,D18S51, D19S433, D21S11, D2S1338, D3S1358, D5S818, D7S820,D8S1179, FGA, TH01, TPOX and vWA) were typed using theAmpFlSTR Identifiler PCR Amplification Kit (Applied Biosystems).One 1.2 mm FTA card punch was mixed with 4 ml AmpFlSTR PCRreaction mix (AB), 2 ml AmpFlSTR Identifiler primer set (AB), 0.2 mlAmpliTaq Gold (AB) and 4 ml ddH2O. The following cycle programmewas used: denaturation and enzyme activation at 95 8C for 11 min,24 cycles of 94 8C for 1 min, 59 8C for 1 min and 72 8C for 1 min,followed by 60 min at 60 8C. PCR reactions were performed in theAB9700 thermal cycler (AB). One ml PCR product was mixed with15 ml Hi-Di formamide (AB) and 0.1 ml GeneScan-500 HD Liz sizestandard (AB), and analysed on an AB Prism 3130XL Genetic Analyzer(AB) with a 36 cm capillary array, POP-4 polymer (AB) and 6 secondinjections at 3000 V. Data were analysed independently by twoanalysts using GeneScan1 analysis software v. 3.7 and Genotyper1

analysis software v. 3.7 (AB), and the results were compared. Theminimum peak height was set to 50 relative fluorescence units(RFUs) for all dyes. All samples were analysed at least twice and theresults from the two experiments were compared. If the resultsdiffered the sample was analysed a third time.

2.4. SNP analysis

A total of 49 autosomal SNPs were amplified by PCR in 25 mlreactions containing one 1.2 mm FTA card punch, 1� PCR buffer,8 mM MgCl2, 700 mM of each dNTP, 0.01–0.17 mM of each primer,and 2 U AmpliTaq Gold DNA polymerase (AB) [16]. The followingcycle programme was used: denaturation at 94 8C for 5 minfollowed by 35 cycles of 95 8C for 30 s, 60 8C for 30 s and 65 8C for30 s followed by 7 min at 65 8C. PCRs were performed in theGeneAmp 9700 (AB) thermal cycler. Excess of PCR primers anddNTPs were removed by addition of a freshly prepared mix of0.75 ml (1 U/ml) shrimp alkaline phosphatase (Amersham Phar-macia Biotech) and 0.023 ml (10 U/ml) Exonuclease I (AmershamPharmacia Biotech) to 2.5 ml PCR product and incubation at 37 8Cfor 1 h followed by incubation at 75 8C for 15 min.

SBE reactions were performed in 8 ml reaction volumes contain-ing 1 ml purified PCR product, 4 ml SNaPshot1 reaction mix (AB),1 ml SBE primer mix (Table S1) and 2 ml Milli-Q water. The SBEprimer mix was diluted in 160 mM ammonium sulphate (Sigma–Aldrich) to minimize primer–dimer artefacts. The SBE reaction wasperformed in GeneAmp 9700 thermal cyclers with 30 cycles of 96 8Cfor 10 s, 50 8C for 5 s, and 60 8C for 30 s. Excess nucleotides wereremoved by addition of 1 ml (1 U/ml) shrimp alkaline phosphatase tothe SBE mix and incubation at 37 8C for 30 min followed by 75 8C for15 min. Two ml SBE products were mixed with 20 ml Hi-Diformamide (AB) and 0.1 ml GeneScan-120 Liz internal size standard(AB). The SBE products were analysed by capillary electrophoresisusing AB Prism 3130XL Genetic Analyzers (AB) with 36 cm capillaryarrays and POP-4 polymer (AB) as previously described [16,21].Analysis was made using GeneScan1 Analysis 3.7 with peakthresholds set to a minimum of 120 RFUs (blue colour), 60 RFUs(green colour), and 30 RFUs (yellow, red, and orange colour). Allelecalls were made independently by two analysts using Genotyper1

C. Børsting et al. / Forensic Science International: Genetics 4 (2009) 34–4236

3.7 macros and the results were compared. If the allele calls differed,the analyses of the electropherograms were repeated by bothanalysts. If the results differed after the second analysis, the resultswere failed and the sample was investigated again. All peaks in thesize standard had to be detected and the peak height of the largestpeak in the electropherogram had to be a minimum of 1000 RFUsbefore analysis could proceed. Peaks detected in pre-defined allelewindows with peak heights larger than 10% (blue colour), 7% (greencolour) or 5% (yellow and red colour) of the maximum peak height inthe respective colour were labelled with allele name, peak heightand size. All samples were analysed at least twice and the resultsfrom the two experiments were compared. If the results differed thesample was analysed a third time.

2.5. Analysis of approved allele calls

Peak heights of all the approved allele calls (24,922 genotypes)were exported to Excel. The data was divided in heterozygote(10,960 genotypes) and homozygote (13,962 genotypes) allele calls.For each SNP locus, the average peak height ratio for heterozygotesand the standard deviation of the peak height ratios weredetermined (Table S2). Similarly, the average peak height ofhomozygote allele calls and the standard deviation of the peakheights were determined (Table S2). From 58 randomly selectedelectropherograms, the peak height of the highest background peakin each dye window was collected and an arbitrary maximumbackground level was determined (R110 dye (blue): 116 RFUs, R6Gdye (green): 120 RFUs, TAMRA dye (yellow): 82 RFU, ROX dye (red):74 RFUs). The background levels were used to calculate an arbitrarysignal/noise ratio for homozygote allele calls (Table S2). Forheterozygote allele calls, the peak height for one allele was plottedagainst the peak height of the other allele (see Fig. 1 for examples) toillustrate the grouping of the observations from each locus.

The acceptable peak height ratio for heterozygote allele calls fora given locus was determined by sorting the peak height ratiosaccording to size, identifying values differing by more than 0.1from the next or previous value on the list, and ignoring all valuesabove or below this value on the list. The number of heterozygoteallele calls per locus varied from 127 to 278, and on average 3.0values (minimum 0 and maximum 22 values) were ignored perlocus data set (Table 1). The acceptable signal/noise ratio forhomozygous allele calls for a given locus was determined by two

Fig. 1. The peak height of one allele was plotted against the peak height of the other al

allele calls). (B) rs251934 (215 allele calls). A total of 15 allele calls for the rs729172 l

identified in (A).

rules: (1) The signal/noise ratio should be at least five, and (2) thesignal/noise ratio for a homozygote allele call should be at least 3times higher than the highest acceptable peak height ratio forheterozygote allele calls in that locus. The estimated minimumpeak height of a homozygous allele call for each of the 98 alleleswas determined by multiplying the smallest acceptable signal/noise ratio with the arbitrary maximum background leveldetermined above (Table 1). This number was rounded to thenearest hundred or rounded up to the nearest thousand.

The acceptable peak height ratio for heterozygote allele callsand the acceptable signal/noise ratio for homozygous allele callsfor each locus (Table 1) were outlined on a single dimensionlogarithmic scale (Fig. S1).

2.6. Determination of allele windows for the Genotyper1 analysis

macro

The nucleotide lengths of all the approved allele calls (24,922genotypes) were exported to Excel. The number of approved allelecalls for a given allele in the data set was between 169 and 491. Theacceptable size range for a given allele (Table S3) was determined bysorting the observations by size, identifying values differing by morethan 0.1 nucleotide (nt) from the next or previous value on the list,and ignoring all values above or below this value on the list. Onaverage, 1.8 values (minimum 0 and maximum 25 values) wereignored per allele data set. The average of the acceptable size rangeswas 0.59 nt (minimum 0.28 nt and maximum 1.48 nt) (Table S3). Theacceptable size range for each allele was used to define the allelewindows in the Genotyper1 analysis macro. In the Genotyper1

analysis macro, the minimum size of an allele window was set to0.6 nt.

Weak signals from extension of PCR products with R6G-labelledddATP [23,24] were observed in the size range where SBE productswere also detected (25–80 nt). The size ranges of the three mostfrequent signals from PCR products were determined (Table S3) asdescribed above for the allele calls, and windows for the extendedPCR products were defined in the Genotyper1 analysis macro.

2.7. Mixture study

DNA from two individuals was isolated from 200 ml blood usingthe QIAamp1 DNA Blood Mini Kit (Qiagen). The purified DNA was

lele for all approved heterozygote allele calls for a given locus. (A) rs729172 (232

ocus were categorized as partially silent alleles (Table 2) and they can be clearly

Table 1Guidelines for allele calling.

Code SNP locus Heterozygous allele call Allele Homozygous allele call

Range Acceptable allele

calls (%)

Signal/noise Estimated minimum

peak height

Acceptable allele

calls (%)

A01 rs1490413 0.6–1.7 99.6 Allele C >10 1000 100.0

Allele T >10 1000 100.0

A02 rs876724 0.8–1.5 100.0 Allele C >10 1000 99.2

Allele T >10 1000 95.7

A03 rs1357617 0.7–2.2 99.4 Allele A >10 1000 100.0

Allele T >10 1000 100.0

A04 rs2046361 1.0–2.0 99.6 Allele A >10 1000 100.0

Allele T >10 1000 100.0

A05 rs717302 2.5–5.0 98.2 Allele G >20 2000 98.4

Allele A >5 600 99.4

A06 rs1029047 0.5–2.0 97.5 Allele A >10 1000 100.0

Allele T >10 1000 98.3

A07 rs917118 0.9–2.0 89.4 Allele G >10 1000 100.0

Allele A >10 1000 100.0

A08 rs763869 0.7–1.5 100.0 Allele C >10 1000 99.4

Allele T >10 1000 99.1

A09 rs1015250 0.8–2.8 96.6 Allele G >10 1000 100.0

Allele C >10 1000 99.4

A10 rs735155 0.3–0.7 99.6 Allele C >5 400 98.9

Allele T >10 700 100.0

A11 rs901398 0.5–1.0 99.5 Allele C >10 1000 100.0

Allele T >10 1000 100.0

A12 rs2107612 1.9–3.6 98.1 Allele G >20 2000 100.0

Allele A >10 1000 99.2

A14 rs1454361 1.1–2.1 98.4 Allele A >10 1000 100.0

Allele T >10 1000 100.0

A15 rs2016276 0.6–1.2 97.5 Allele C >10 1000 100.0

Allele T >10 1000 100.0

A16 rs729172 2.0–6.0 93.1 Allele G >20 2000 98.9

Allele T >10 800 97.6

A17 rs740910 2.0–4.0 99.1 Allele G >20 2000 100.0

Allele A >10 1000 100.0

A18 rs1493232 0.6–1.7 100.0 Allele G >10 1000 100.0

Allele T >10 1000 100.0

A19 rs719366 1.0–2.0 98.8 Allele C >10 1000 100.0

Allele T >10 1000 98.3

A20 rs1031825 2.5–5.0 99.1 Allele G >20 2000 100.0

Allele T >5 400 100.0

A21 rs722098 2.0–3.6 95.2 Allele G >20 2000 100.0

Allele A >10 1000 99.7

A22 rs733164 2.0–6.0 100.0 Allele G >20 2000 99.6

Allele A >10 1000 100.0

A23 rs826472 1.0–2.2 97.9 Allele G >10 1000 100.0

Allele A >10 1000 100.0

A24 rs2831700 0.9–2.1 99.6 Allele C >10 1000 100.0

Allele T >10 1000 98.3

A25 rs873196 2.5–5.0 99.5 Allele G >20 2000 85.4

Allele A >5 600 88.9

A26 rs1382387 2.5–5.1 100.0 Allele G >20 2000 97.9

Allele T >5 400 100.0

A27 rs2111980 0.4–1.0 100.0 Allele G >10 1000 100.0

Allele A >10 1000 100.0

A28 rs2056277 0.8–2.2 100.0 Allele G >10 1000 100.0

Allele A >10 1000 96.8

A29 rs1024116 1.1–2.8 100.0 Allele G >10 1000 98.3

Allele A >5 600 98.7

C. Børsting et al. / Forensic Science International: Genetics 4 (2009) 34–42 37

Table 1 (Continued )

Code SNP locus Heterozygous allele call Allele Homozygous allele call

Range Acceptable allele

calls (%)

Signal/noise Estimated minimum

peak height

Acceptable allele

calls (%)

A30 rs727811 1.0–2.0 99.2 Allele A >10 1000 99.3

Allele C >10 800 97.0

A32 rs1413212 0.6–1.2 100.0 Allele C >5 400 97.5

Allele T >10 700 98.3

A33 rs938283 0.5–1.1 100.0 Allele C >10 1000 100.0

Allele T >10 1000 100.0

A34 rs1979255 2.4–4.5 98.8 Allele G >20 2000 99.0

Allele C >10 800 100.0

A36 rs2076848 0.6–2.5 97.1 Allele A >10 1000 100.0

Allele T >10 700 98.1

A37 rs1355366 0.7–1.3 99.3 Allele C >10 800 99.0

Allele T >10 700 99.3

A38 rs907100 0.6–2.4 97.8 Allele G >10 1000 98.8

Allele C >10 800 99.3

A39 rs354439 1.0–1.7 99.2 Allele A >10 1000 90.9

Allele T >10 700 95.5

A40 rs2040411 1.0–2.0 97.9 Allele G >10 1000 100.0

Allele A >5 600 98.9

A41 rs737681 0.4–1.1 98.7 Allele C >10 800 97.2

Allele T >10 700 100.0

A42 rs2830795 0.9–2.1 97.9 Allele G >10 1000 100.0

Allele A >5 600 97.1

A43 rs251934 1.2–1.9 99.1 Allele G >10 1000 100.0

Allele A >10 1000 100.0

A44 rs914165 0.3–1.1 98.8 Allele C >5 400 97.1

Allele T >10 700 94.5

A45 rs10495407 0.4–0.8 98.9 Allele C >5 400 95.9

Allele T >10 700 97.9

A46 rs1360288 1.7–3.5 98.7 Allele G >20 2000 97.8

Allele A >5 600 100.0

A48 rs964681 0.5–1.3 100.0 Allele C >10 1000 99.1

Allele T >10 1000 100.0

A49 rs1005533 1.0–3.0 96.6 Allele G >10 1000 97.2

Allele A >5 600 98.2

A50 rs8037429 1.1–2.0 99.6 Allele G >10 1000 97.9

Allele A >5 600 94.6

A51 rs891700 1.2–3.0 97.7 Allele G >10 1000 100.0

Allele A >5 600 96.7

A52 rs1335873 0.9–2.1 97.4 Allele A >10 1000 85.6

Allele T >10 700 91.7

A54 rs1528460 1.9–4.4 99.2 Allele G >20 2000 98.2

Allele A >5 600 92.5

C. Børsting et al. / Forensic Science International: Genetics 4 (2009) 34–4238

quantified by the Quantifiler1 Human DNA quantification kit (AB)and the samples were diluted in water to a concentration of500 pg/ml. The DNA from the two individuals were mixed in thefollowing ratios: 10:1, 5:1, 2:1, 1:1, 1:2, 1:5, and 1:10. Onenanogram of each mixture was used for PCR. Each mixture wastyped 3 times and analysed with GeneScan1 3.7 and Genotyper1

3.7 (AB) as described above. The peak heights of the detectedalleles were exported from Genotyper1 3.7 and analysed in Excelusing the guidelines developed for allele calling (Table 1).

2.8. Calculation of likelihood ratio (LR)

Paternity indices and sibling indices (SIs) were calculated usingDNAVIEW 27.21 and twin indices (TIs) were calculated in Excel(Table 3).

PI = P (observed genotypesjH)/P (observed genotypesjHA),where the hypothesis H was ‘‘the tested man is the father’’, andthe alternative hypothesis HA was ‘‘a random man is the father’’.

SI = P (observed genotypesjH)/P (observed genotypesjHA),where the hypothesis H was ‘‘the tested children are full siblings’’,and the alternative hypothesis HA was ‘‘the tested children areunrelated’’.

TI = P (observed genotypesjH)/P (observed genotypesjHA),where the hypothesis H was ‘‘the twins are monozygotic twins’’,and the alternative hypothesis HA was ‘‘the twins are dizygotictwins’’.

Frequency databases for Danes were used for all calculations[16,19]. Calculations were performed under the assumption thatsilent alleles exist. For a locus where a silent allele had notpreviously been observed, the probability of a silent allele was set

Table 2Approved allele calls with unusual peak heights or signal to

noise ratios.

Category Observations

Weak signala 76

Partially silent alleleb 30

Backgroundc 19

Saturated signal 1

Undetermined 17

a Maximum peak height < 3500 RFUs in the R110 dye.b A total of 15 out of 30 found in rs729172.c Caused by poor EXO I degradation of PCR primers.

C. Børsting et al. / Forensic Science International: Genetics 4 (2009) 34–42 39

to 1/(n + 1) (n is the total number of observed alleles). Only twoSNP loci per chromosome, positioned at each end of thechromosomes, were used for calculation of SI and TI in order toensure that the loci used in the calculation were independentlysegregated during meiosis. Typical LRs were calculated as thegeometric means.

3. Results

3.1. Modifications of the SBE primer design

The 52 SNPs in the SNPforID multiplex were originally analysedin two SBE reactions with 23 and 29 SBE primers, respectively. Thelengths of the SBE primers varied from 16 to 92 nucleotides [16].This SBE primer design was altered for three reasons. First, areduction in the average length of the SBE primers was desirable,because long oligonucleotides are expensive and difficult tosynthesize. Second, single base extension of PCR primers leftoverfrom the PCR reaction and not degraded by Exonuclease Ioccasionally resulted in peaks in the 20–30 nucleotide size rangeof the electropherogram and made it difficult to analyse SBEproducts of similar sizes. Third, weak signals from extension of PCRproducts with R6G-labelled ddATP [23,24] were sometimesdetected and may be confused with signals from true alleles.

In the new SBE primer design, the number of loci detected in the20–30 nucleotide size range of the electropherogram was reducedfrom 7 to 3, and overlaps between signals from extended SBE primersand ddATP-extended PCR products were avoided (Table S3). Thelengths of 23 SBE primers were altered in the new design (Table S1),butthetargetsequenceswerethesameasthosepreviouslypublished[16]. Inthe process of making the new design, three SNPs (rs1886510,rs1463729, rs1028528) were excluded from the assay. Two of thethree excluded SNPs (rs1886510 and rs1028528) were previouslyknown to be difficult to analyse and the cause of many extraexperiments, because the signals from the SBE products were weak.The SNP rs1463729 was excluded because the peak height of the Gallele sometimes was 8 times higher than the peak height of the Aallele. This was an unacceptable peak height ratio because the A alleleof a heterozygous individual may not always be detected in weaklyamplified samples. Inthe newdesign,49SNPsaredetected in two SBEreactions with 21 and 28 SBE primers, respectively. The lengths of theSBE primers vary from 16 to 84 nucleotides.

3.2. Definition of guidelines for allele calling

A total of 216 samples from 50 paternity cases and 33 twincases were typed for 49 SNPs. All samples were typed at least twiceand all electropherograms were analysed independently by twoanalysts. Of the 432 profiles generated from the 216 samples typedin duplicate, 339 (78%) were complete profiles. After a third roundof experiments, approved allele calls for all 49 SNPs in twoindependent experiments were obtained for all samples exceptfour. These samples had to be typed a fourth time in order to obtaincomplete SNP profiles.

Peak heights and nucleotide lengths of 24,922 approvedgenotypes were collected in Excel and guidelines for allele calling(Table 1 and Fig. S1) were defined as described in Section 2.Furthermore, the data was used to define allele windows in theGenoTyper analysis macro for all known alleles and windows for themost frequently observed ddATP-extended PCR products (Table S3).

3.3. Troubleshooting by using the guidelines for allele calling

Of the 24,922 genotypes that were approved by manualanalysis, 156 failed the guidelines for heterozygote allele callsand 231 failed the guidelines for homozygote allele calls (Table 1).

The allele calls that differed most from the guidelines (83heterozygote allele calls and 60 homozygote allele calls) wereexamined again and the reasons for the unusual peak height ratioswere categorized into five categories (Table 2).

The majority of these allele calls (76 out of 143) were found insamples where the fluorescent signals from the SBE products wererelatively low. Usually, the strongest fluorescent signal from SBEprimers extended with R110-labelled ddGTP (blue colour) washigher than 8000 RFUs, but in samples where the highest peak wasless than 3500 RFUs, the guidelines for allele calling were often notfulfilled in several loci.

Reproducible heterozygote allele calls with unusual peakheight ratios were categorized as partially silent alleles(Table 2). In some cases, the trait was also inherited and foundin one of the parents and the off-spring, or in twins. Half of theallele calls categorized as partially silent alleles (15 out of 30) werefound in seven samples in the locus rs729172. The normal peakheight ratio for the G/T SNP rs729172 was approximately 4(Table 1), but for the seven samples, the average ratio was 0.75. Inthe heterozygote plot shown in Fig. 1A, the 15 unusual peak heightratios could be clearly distinguished from the majority ofheterozygote allele calls found in rs729172. DNA from selectedsamples was amplified using another set of PCR primers forrs729172, and sequencing of the PCR products confirmed that theseven samples were heterozygous G/T at rs729172 (data notshown). In addition, a previously unknown C/T polymorphism wasfound in position �15 of the reverse PCR primer in all sevensamples, whereas sequencing of control samples showed theexpected C in this position. The reverse PCR primer for rs729172was redesigned with a degenerated nucleotide in position�15, andwith the new PCR primer, the average peak height ratio in theseven samples changed from 0.75 to 3.8 (data not shown). Thisstrongly indicates that a T complementary to the nucleotide inposition �15 of the reverse PCR primer caused poor amplificationof the rs729172 G allele. The other 15 allele calls categorized aspartially silent alleles were found in six different loci in sevendifferent individuals. These partially silent alleles have not beeninvestigated further.

A total of 19 allele calls failed to fulfil the guidelines for allelecalling because of high background in the 20–30 nucleotides sizerange of the electropherogram (Table 2). All 19 allele calls werefound in the same experiment and the background was probablycaused by single base extension of poorly degraded PCR primers.During this work, the mixture of Shrimp alkaline phosphatase andExonuclease I, which is used to degrade dNTPs and PCR primersleftover from the PCR reaction, was stored at 4 8C for up to oneweek. However, it was later found that the background in the 20–30 nucleotides size range of the electropherogram was stronglyreduced if a freshly prepared mixture of Shrimp alkalinephosphatase and Exonuclease I was used.

The 387 allele calls that were approved by manual analysis butfailed the guidelines for allele calling were correct, except for onehomozygous allele call in rs914165. The individual was hetero-

Fig. 2. DNA from two individuals was mixed in seven different ratios (10:1, 5:1, 2:1, 1:1, 1:2, 1:5, and 1:10) and typed for 49 SNPs. The results from the rs917118 locus (A07)

are shown. At the top, the SBE products extended with R110-labelled ddGTP are shown, and below, the SBE products extended with R6G-labelled ddATP are shown. The locus

code and allele call (A07G or A07A), and the peak height in RFUs are shown in boxes below the peak. The two individuals were opposite homozygotes in the locus rs917118.

The two alleles were clearly detected in all mixtures. According to the guidelines in Table 1, the acceptable peak height ratio for heterozygote allele calls for the rs917118 locus

was 0.9–2.0. Only the 1:1 and the 2:1 mixtures fell inside this range.

C. Børsting et al. / Forensic Science International: Genetics 4 (2009) 34–4240

zygous C/T in rs914165, but in one experiment, the individual wascalled as homozygous T. The amplification of the sample was weak(the highest peak in the R110 dye was less than 3500 RFUs, seeabove), and the analysts failed to detect the C allele. The individualwas typed 3 times and in the two experiments where theindividual was called as a heterozygote, the C allele of the rs914165locus could be categorized as a partially silent allele (data notshown). Interestingly, the sample belonged to a child, and themother of the child also had a weakly amplified C allele.

3.4. Detection of mixtures using the guidelines for allele calling

Purified DNA from two individuals was quantified and mixed inseven different ratios (10:1, 5:1, 2:1, 1:1, 1:2, 1:5, and 1:10). Themixtures were typed for the 49 SNPs and the electropherogramswere analysed using the guidelines in Table 1.

The peak height ratios of many loci in each mixture wereimmediately called as suspicious and all mixtures were easilyidentified as mixtures by SNP typing. In the five loci where the

Fig. 3. DNA from two individuals was mixed in seven different ratios (10:1, 5:1, 2:1, 1:1, 1

are shown. At the top, the SBE products extended with TAMRA-labelled ddCTP are show

locus code and allele call (A37C or A37T), and the peak height in RFUs are shown in boxes

whereas the other individual was homozygous C. The two alleles were clearly detected in

for heterozygote allele calls for the rs1355366 locus was 0.7–1.3. Only the 1:10 mixtu

two individuals were opposite homozygotes, the two alleleswere clearly visible in all mixtures (see example in Fig. 2) exceptfor the rs2056277 locus, where only the G allele was detected ina 1:10 mixture. As expected, the peak height ratio of a 1:1mixture of opposite homozygotes fell inside the acceptablerange for heterozygote allele calls according to the guidelines inTable 1. Peak height ratios in most of the 1:2 mixtures ofopposite homozygotes fell just outside the acceptable range forheterozygote allele calls, whereas peak height ratios in all 1:5 or1:10 mixtures of opposite homozygotes were clearly differentfrom the guidelines and the genotypes were not accepted by theanalysts.

In 22 of the 49 loci, one individual was heterozygous and theother individual was homozygous (see example in Fig. 3). At theseloci, most peak height ratios fell inside the acceptable range forheterozygote allele calls when DNA from the individual with aheterozygous genotype was present in 5 or 10 times higherconcentration than DNA from the individual with a homozygousgenotype. However, in 1:1 mixtures or when DNA from the

:2, 1:5, and 1:10) and typed for 49 SNPs. The results from the rs1355366 locus (A37)

n, and below, the SBE products extended with ROX-labelled ddTTP are shown. The

below the peak. One of the individuals was heterozygous C/T in the locus rs1355366,

all mixtures. According to the guidelines in Table 1, the acceptable peak height ratio

res fell inside this range.

Table 3PIs from 43 true trios, SIs from 13 full siblings, and TIs from 20 monozygotic twins.

Paternity indices Sibling indices Twin indices

SNPs STRs SNPs STRs SNPs STRs

Typical LR 5.07E+05 1.55E+07 1.43E+03 2.36E+04 5.90E+08 3.11E+06

Minimum 560 270 0.1 5.7 1.62E+08 1.17E+06

Maximum 2.33E+09 3.45E+10 3.51E+06 2.48E+09 2.20E+09 6.41E+06

C. Børsting et al. / Forensic Science International: Genetics 4 (2009) 34–42 41

individual with a homozygous genotype was present in the highestconcentration, the peak height ratios were clearly different fromthe guidelines and the genotypes were not accepted.

3.5. Casework analyses

In addition to the 49 SNPs, all samples were typed for 15 STRswith the AmpFlSTR Identifiler PCR Amplification Kit as part of thekinship investigation. The conclusions based on STRs and SNPswere concordant in all cases.

A total of 43 of the 50 paternity cases were true trios and thepaternity indices were calculated (Table 3). The typical PI based onSTRs was approximately 30 times higher than the typical PI basedon SNPs, which was comparable to the results obtained previously[19]. In one of the true trios, mother, child, and tested man had thegenotypes 14;16, 14;19, and 15;18, respectively, in the STR locusvWA. This mismatch was considered a single-step mutationbetween the tested man and the child. The PI based on the 15 STRsin this case was only 270, whereas the PI based on 49 SNPs was520,000. No mutation was observed in any of the SNP loci in any ofthe true trios. In the seven paternity cases, where the tested manwas unrelated, the average number of mismatches between thetested man and the child was 10.3 for the STR investigation(minimum 8, maximum 12) and 8.1 for the SNP investigation(minimum 5, maximum 12).

The twin index was calculated in the 20 twin cases where thetwins had identical profiles (Table 3). The typical TI based on SNPswas approximately 100 times higher than that obtained with STRs.In the remaining twin cases, the twins were not identical and asibling index was calculated (Table 3). The typical SI based on STRswas approximately 10 times higher than that obtained with SNPs.In the 13 twin cases, where the twins were dizygotic, the averagenumber of discordant loci was 9.2 for the STR investigation(minimum 5, maximum 13) and 18.0 for the SNP investigation(minimum 12, maximum 26).

4. Discussion

The guidelines for allele calling developed in this work werebased on the data generated from independent analyses by twoexpert analysts and the duplicate typing of 216 samples. Peakheights and nucleotide lengths from 35,882 allele calls (10,960heterozygote and 13,962 homozygote genotypes) were collected.For each locus, the variation in peak height ratios was analysed. Forheterozygous allele calls, the peak height of one allele was dividedby the peak height of the other allele. For homozygous allele calls,the peak height was divided by an average maximum noise level.Two simple criteria were set up to define the guidelines for eachlocus: (1) the peak height should be at least 5 times higher than thenoise, and (2) the signal to noise ratio for homozygous allele callsshould be at least 3 times higher than the highest acceptable peakheight ratio for heterozygous allele calls. The two criteria wereused to ensure that only strong signals would be accepted ashomozygous allele calls and that there was a clear distinctionbetween the peak height ratios of homozygous and heterozygousallele calls. Furthermore, the criteria ensured that heterozygote

allele calls with unusual peak height ratios would not be acceptedeven though both alleles were clearly detected. For most SNP loci,the variation in peak height ratios for heterozygote allele calls wassmall, and the acceptable signal to noise ratio for homozygousallele calls was defined to be more than 3 times higher than thehighest acceptable peak height ratio for heterozygote allele calls.

The peak height ratios of 387 allele calls (1.5%) were consideredunusual (see Section 2) and did not fulfil the guidelines forheterozygote and homozygote allele calls. The allele calls thatdiffered most from the guidelines were scrutinized, and themajority of them were detected in samples that were either weaklyamplified, or in electropherograms where the noise was unusuallyhigh. All except one of these allele calls were correctly assigned,but they failed to fulfil the guidelines, because of typical problemsthat would normally require an expert analyst to evaluate theresult. Thus, the guidelines seemed to divide the allele calls intotwo groups: those that were simple to analyse, and those that weredifficult to analyse or should be rejected.

The laboratory protocol developed for analysis of the SBEproducts were divided into four steps: (1) analysis of the CE datafiles using analysis software (typically Genotyper1 or GeneMap-per1), where all relevant peaks were labelled with locus, allele,length, and peak height, (2) export of a text file containing the loci,alleles, and peak heights of the allele calls, (3) import of the text filein Excel, where the peak height ratios were matched against theguidelines for allele calling using simple Excel macros, and finally(4) re-analysis of the allele calls that did not fulfil the guidelines,where the expert analyst must decide whether the allele callshould be approved or rejected. Obviously, analysis of theelectropherograms would be more efficient if the guidelines forallele calling could be defined as analysis parameters in theanalysis software. However, this is not possible with the currentversions of the softwares.

The guidelines for allele calling and the protocol for SBE analysisdescribed here proved to be a highly efficient method for detectionof mixtures or contaminated DNA preparations. This is animportant observation because samples collected from crimescenes may contain genomic DNA from more than one individual.Mixtures may easily be detected using the conventional STR kits,because more than two alleles are often detectable in several loci.However, most SNP loci are bi-allelic, and concerns have beenraised whether mixtures may be identified by SNP typing. Theresults shown here demonstrated that the peak height ratios ofnumerous allele calls in a mixture differed significantly from theexpected ratios and this will clearly identify the DNA mixtures.Nevertheless, more work needs to be done in this field because theSBE assay most likely will be used on highly degraded crime scenesamples, and the peak height ratios in these samples may differfrom peak height ratios generated from reference samples.

The SBE assay with 49 of the 52 SNPs from the SNPforIDmultiplex was accredited for forensic genetic testing according tothe ISO 17025 standard in the fall of 2007. The assay is used as asupplement to STR typing in all paternity and immigration casesthat do not involve trios (woman, child, man) and in cases, whereone or two mismatches are detected between the tested man andthe child, or between the tested woman and the child. In the first 70

C. Børsting et al. / Forensic Science International: Genetics 4 (2009) 34–4242

cases where SNPs were reported under the ISO 17025 accredita-tion, 316 samples were typed in duplicate and complete SNPprofiles were obtained for all individuals. No mutations weredetected in any of the SNP loci and only four of the 15,484 reportedgenotypes were unusual. Two unrelated individuals from Africahad two null alleles in rs2830795, and reproducible heterozygoteallele calls with unusual peak height ratios (partially silent alleles)were detected in rs737681 and rs938283 in two individuals ofnon-European origin. An analysis of all the SBE experiments in the70 cases showed that complete SNP profiles were generated in 76%of the experiments. In 12% of the experiments, the results wererejected in one SNP locus, and in 5% of the experiments, the resultswere rejected in two SNP loci. In the remaining 7% of theexperiments, the results were rejected in more than two loci.Interestingly, the number of complete SNP profiles increased from70% in the first 20 cases to 84% in the last 20 cases, which indicatesthat the performance of the assay has improved with increasinglaboratory experience (data not shown).

In a typical paternity case with mother, child and tested man,there are two types of mismatches to consider at a bi-allelic SNPlocus [19]: (1) genetic inconsistency, where the child is hetero-zygous, and the mother and the tested man are homozygous for thesame allele, and (2) opposite homozygosity, where the child andthe tested man are homozygous for different alleles. For calculationof likelihood ratios, the mutation probability, mSNP, was set to 10�6

and the probability of silent alleles was set to 1/(n + 1), where n isthe total number of observed alleles in the database (n is currently371 for Danes). Thus, with a mutation probability ofmSTR = 3 � 10�3 for a typical STR locus, one genetic inconsistencyin a SNP locus count against paternity with approximately 10 timesmore weight than two genetic inconsistencies in STR loci.Furthermore, opposite homozygosity between the tested manand the child at three SNP loci counts against paternity with at leasttwice the weight of genetic inconsistencies in two STR loci. Thisillustrates how powerful the SNPs are in paternity testing and howuseful the SNPs are in cases, where a mutation is suspected in anSTR locus.

Six of the 49 SNPs are positioned on the same chromosome armas another SNP in the multiplex [16,19] and with only 1–4recombination events per chromosome per meiosis [25,26], theseSNP pairs will occasionally be inherited as haplotypes and not asindependent loci. This may be ignored in the calculation of PIs intrios and duos, because the loci are in linkage equilibrium in thepopulation and because the haplotypes of the parents areunknown [19]. However, in kinship cases with more complexpedigrees, the loci must be independent in order to use the productrule for calculation of likelihood ratios. Therefore, it is prudent touse only two SNP loci per chromosome for the calculation ofindices in these cases until the statistics of closely positionedmarkers in kinship testing has been further evaluated.

To our knowledge, this is the first time a multiplex SNP typingassay of this size has been introduced in routine forensic geneticcasework and has been accredited according to the ISO 17025standard. We anticipate that SNPs will become an important partof the forensic arsenal in the future and that additional SNP typingassays will follow, e.g. assays for particular physical traits or ethnicorigin.

Acknowledgements

We thank Juan J. Sanchez, PhD for helpful discussions, and wethank Nadia Jochumsen and Annemette H. Birk for technicalassistance.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in

the online version, at doi:10.1016/j.fsigen.2009.04.004.

References

[1] D.J. Reeder, Impact of DNA typing on standards and practice in the forensiccommunity, Arch. Pathol. Lab. Med. 123 (1999) 1063–1065.

[2] P.M. Schneider, Scientific standards for studies in forensic genetics, Forensic Sci.Int. 165 (2007) 238–243.

[3] N. Morling, R.W. Allen, A. Carrecedo, H. Geada, F. Guidet, C. Hallenberg, W. Martin,W.R. Mayr, B. Olaisen, V.L. Pascali, P.M. Schneider, Paternity testing commission ofthe International Society of Forensic Genetics: recommendation on geneticinvestigation in paternity cases, Forensic Sci. Int. 129 (2002) 148–157.

[4] ENFSI standing committee for quality and competence, Standards for accredita-tion, www.enfsi.org/documents/policy documents/QCC-ACC-001.pdf.

[5] E. Malkoc, W. Neuteboom, The current status of forensic science laboratoryaccreditation in Europe, Forensic Sci. Int. 167 (2007) 121–126.

[6] M.J. Saks, J.J. Koehler, The coming paradigm shift in forensic identification science,Science 309 (2005) 892–895.

[7] A. Opar, Crime and punishment, Nature Med. 12 (2006) 1110–1112.[8] H. Whittall, The forensic use of DNA: scientific success story, ethical minefield,

Biotechnol. J. 3 (2008) 303–305.[9] M. Krawczak, Informativity assessment for biallelic single nucleotide polymorph-

isms, Electrophoresis 20 (1999) 1676–1681.[10] P. Gill, An assessment of the utility of single nucleotide polymorphisms (SNPs) for

forensic purposes, Int. J. Legal Med. 114 (2001) 204–210.[11] K.L. Ayres, The expected performance of single nucleotide polymorphism loci in

paternity testing, Forensic Sci. Int. 154 (2005) 167–172.[12] S. Inagaki, Y. Yamamoto, Y. Doi, T. Takata, T. Ishikawa, K. Imabayashi, K. Yoshi-

tome, S. Miyaishi, H. Ishizu, A new 39-plex analysis method for SNPs including 15blood group loci, Forensic Sci. Int. 144 (2004) 45–57.

[13] E. Petkovski, C. Keyser-Tracqui, R. Hienne, B. Ludes, SNPs and MALDI-TOF MS:tools for DNA typing in forensic paternity testing and anthropology, J. Forensic.Sci. 50 (2005) 1–7.

[14] H.Y. Lee, M.J. Park, J. Yoo, U. Chung, G. Han, K. Shin, Selection of twenty-four highlyinformative SNP markers for human identification and paternity analysis inKoreans, Forensic Sci. Int. 148 (2005) 107–112.

[15] L.A. Dixon, C.M. Murray, E.J. Archer, A.E. Dobbins, P. Koumi, P. Gill, Validation of a21-locus autosomal SNP multiplex for forensic identification purposes, ForensicSci. Int. 154 (2005) 62–77.

[16] J.J. Sanchez, C. Phillips, C. Børsting, K. Balogh, M. Bogus, M. Fondevila, C.D.Harrison, E. Musgrave-Brown, A. Salas, D. Syndercombe-Court, P.M. Schneider,A. Carracedo, N. Morling, A multiplex assay with 52 single nucleotide polymorph-isms for human identification, Electrophoresis 27 (2006) 1713–1724.

[17] L. Li, C. Li, R. Li, Y. Liu, Y. Lin, T. Que, M. Sun, Y. Li, SNP genotyping by multiplexamplification and microarrays assay for forensic identification, Forensic Sci. Int.162 (2006) 74–79.

[18] A.J. Pakstis, W.C. Speed, J.R. Kidd, K.K. Kidd, Candidate SNPs for a universalindividual identification panel, Hum. Genet. 121 (2007) 305–317.

[19] C. Børsting, J.J. Sanchez, H.E. Hansen, A.J. Hansen, H.Q. Bruun, N. Morling, Per-formance of the SNPforID 52 SNP-plex assay in paternity testing, Forensic Sci. Int.Genet. 2 (2008) 292–300.

[20] E. Musgrave-Brown, D. Ballard, K. Balogh, K. Bender, B. Berger, M. Bogus, C. Børsting,M. Brion, M. Fondevila, C. Harrison, C. Oguzturun, W. Parson, C. Phillips, C. Proff, E.Ramos-Luis, J.J. Sanchez, P.S. Diz, B. Rey, B. Stradmann-Bellinghausen, C. Thacker, A.Carracedo, N. Morling, R. Scheithauer, P.M. Schneider, D.S. Court, Forensic validationof the SNPforID 52-plex assay, Forensic Sci. Int. Genet. 1 (2007) 186–190.

[21] J.J. Sanchez, C. Børsting, K. Balogh, B. Berger, M. Bogus, J. Butler, A. Carracedo, D.Syndercombe Court, L.A. Dixon, B. Filipovic, M. Fondevila, P. Gill, C.D. Harrison, C.Hohoff, R. Huel, B. Ludes, W. Parson, T.J. Parsons, C. Phillips, H. Schmitter, P.M.Schneider, P.M. Vallone, N. Morling, Forensic typing of autosomal SNPs with a 29SNP-multiplex—results of a collaborative EDNAP exercise, Forensic Sci. Int. Genet.2 (2008) 176–183.

[22] A.J. Hansen, B.T. Simonsen, C. Børsting, C. Hallenberg, N. Morling, Semi-automaticpreparation of biological database samples for STR and SNP typing, Prog. For.Genet. 11, International Congress Series 1288 (2006) 663–665.

[23] G. Hu, DNA polymerase-catalyzed addition of nontemplated extra nucleotides tothe 3’ end of a DNA fragment, DNA Cell Biol. 12 (1993) 763–770.

[24] V.L. Magnuson, D.S. Ally, S.J. Nylund, Z.E. Karanjawala, J.B. Rayman, J.I. Knapp, A.L.Lowe, S. Ghosh, F.S. Collins, Substrate nucleotide-determined non-templatedaddition of adenine by Taq DNA polymerase: implications for PCR-based geno-typing and cloning, Biotechniques 21 (1996) 700–709.

[25] A. Kong, D.F. Gudbjartsson, J. Sainz, G.M. Jonsdottir, S.A. Gudjonsson, B. Richards-son, S. Sigurdardottir, J. Barnard, B. Hallbeck, G. Masson, A. Shlien, S.T. Palsson,M.L. Frigge, T.E. Thorgeirsson, J.R. Gulcher, K. Stefansson, A high-resolutionrecombination map of the human genome, Nat. Genet. 31 (2002) 241–247.

[26] F. Sun, M. Oliver-Bonet, T. Liehr, H. Starke, E. Ko, A. Rademaker, J. Navarro, J. Benet,R.H. Martin, Human male recombination maps for individual chromosomes, Am.J. Hum. Genet. 74 (2004) 521–531.