University of Groningen
Finding causal variants for complex genetic disease
Spijker, Geert Theodoor
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to
cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2007
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Spijker, G. T. (2007). Finding causal variants for complex genetic disease: the contribution of statistical
methodology to fine-mapping and assay optimization s.n.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the
author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the
number of authors shown on this cover page is limited to 10 maximum.
Download date: 15-06-2017
23
A factorial experiment for optimizing the PCR
conditions in routine genotyping*
Marijke Niens1, Geert T. Spijker1, Arjan Diepstra2 and Gerard J. te Meerman1
Departments of 1Medical Genetics and 2Pathology and Laboratory Medicine, University Medical Centre
Groningen, the Netherlands
Although most PCRs would produce proper PCR products when first tried, a general
optimization is required to yield the best results. This optimization is often achieved
by changing one factor at a time. However, this may lead to suboptimal results, since
interactions between conditions are difficult to detect with this approach. In the
present study, we describe the factorial optimization of PCR conditions for
microsatellite genotyping, by introducing small systematic variations in conditions
during the genotyping process. The hypothesis was that small changes will not affect
genotyping results, but will provide information about the optimality of current
conditions. The conditions to vary were the concentrations of buffer, MgCl2, dNTPs,
primers, Taq polymerase and DNA, the annealing temperature (Ta) and the number
of cycles. We show that, by a 28 factorial experiment, it is possible to identify not only
the factors responsible for obtaining good results, but also those responsible for bad
results. However, the condition leading to the highest signals is not necessarily the
best operational condition. The best operational condition is minimally sensitive to
random pipetting fluctuations and yields reliable genotypes as well.
Introduction
High-throughput genotyping is a core technology for many types of genetic
investigations. In linkage and association studies, hundreds to thousands of
individuals are genotyped for thousands of different markers. Nowadays, this is
done in a semi-automated way using PCRs prepared with pipetting robots and
using 384-well thermal cyclers. The subsequent steps consist of capillary
Reprinted, with permission, from:
Biotechnology and Applied Biochemistry (2005; 42: 157-62), Portland press Ltd.
*
24
Chapter 2
electrophoresis or microarray technology. This is a high-throughput, almost
industrial, process in which it is crucial to maintain reliable process conditions. The
first-pass genotyping results have to be as complete as possible, to reduce high costs
and workload due to laborious rework after failure.
Insufficient quantity of PCR products is a crucial item in the genotyping process [1].
Optimization of PCR is often tried by sequential optimization of each reaction
variable, an approach that rarely leads to complete testing of all possible
combinations. In practice, the optimum conditions are rarely identified [2]. It so
happens that routine genotyping results are below those obtained during
optimization. Such failures are often interpreted as a result of random errors that are
difficult to control and identify. While there is some truth in this, it might also mean
that sequential optimization of PCR conditions may not be sufficient to obtain a
robust working point. Production conditions should be minimally sensitive to
uncontrollable and unavoidable random fluctuations, often related to pipetting
variation.
In a previous study, we showed that, by use of ANOVA, applied to the first-pass
results of a genetic study, important factors influencing genotyping success can be
determined [3]. Once the potential strength of ANOVA has been understood, we
introduced small controlled and designed experimental variations into the PCR
conditions of the genotyping process to obtain information about the optimality of
our current working conditions and sensitivity of those conditions to random
changes.
We propose to view optimization of PCR conditions as a continuing process, which
does not stop after initial optimization. We applied small systematic changes (in the
order of magnitude of random pipetting errors or less) to the best estimate of
optimal process conditions. By systematic combination of the conditions, production
runs can be set up as factorial experiments. We show data where a 28 factorial
design was used, in which each factor assumes two levels. In this design, all
combinations of factors occur equally frequently. This type of design is able to
estimate the effect of each PCR factor and can check possible interactions between
the factors [4]. Since there are complex interactions among the individual variables
in the PCR [5], one set of amplification conditions that is optimal for all situations
may not exist. The experiments were performed on an existing genotyping project
[6], because it is expensive and rather inefficient to generate data only to improve
the genotyping process and because applying systematic small changes to process
conditions is not likely to affect the success of genotyping. We report here the results
of one large and one smaller experiment, where 384 individuals are genotyped with
27 separate microsatellite markers, using fluorescence strength as the dependent
variable and process conditions as experimental factors.
Optimizing PCR conditions
25
Table 1: PCR conditions
Factor
-10% of standard
Standard condition
+10% of standard
DNA
22.5 ng
25 ng
27.5 ng
- Tris/HCl
9 mM
10 mM
11 mM
- KCl
45 mM
50 mM
55 mM
1.35 mM
1.5 mM
1.65 mM
MgCl2*
0.9 mM
1.0 mM
1.1 mM
dNTPs
0.18 mM
0.2 mM
0.22 mM
Buffer
- MgCl2
Each primer
0.225 μM
0.25 μM
0.275 μM
Taq polymerase*
0.45 units
0.5 units
0.55 units
Ta* (experiment 1)
-2 °C
Primer specific Ta
+2 °C
(experiment 2)
-1 °C
+1 °C
Number of PCR cycles
30
32
34
Standard PCR conditions and experimental conditions of experiment 1, in a final volume of 10 μl. Asterisks (*)
mark factors that were also tested in experiment 2.
Materials and methods
Source of template
We performed experiments within a research project where it was required to
genotype markers covering the HLA (human leucocyte antigen) region in a genetic
association study for Hodgkin lymphoma [6]. Leucocyte (germline) DNA was
extracted from 20 ml of EDTA/blood by standard procedures [7] and stored at −80
°C. Before use, the DNA was diluted in MilliQ water and stored wet at −20 °C.
PCR and genotyping procedure
The PCR contained the PCR buffer (100 mM Tris/HCl, pH 9.0, 500 mM KCl and 15
mM MgCl2), Taq DNA polymerase (Amersham Biosciences, Uppsala, Sweden),
MgCl2, dNTPs (Roche Diagnostics, Mannheim, Germany) and primers [with one 5’labelled with fluorochrome 6-Fam (6-carboxyfluorescein) or HEX
(hexachlorofluorescein); Sigma,Malden, The Netherlands]. The concentrations of the
basic amplification reaction mixture for the two experiments are shown in Table 1.
Reactions were carried out in a final volume of 10 μl. In experiment 1, genotyping
was performed on 384 samples in a 384-well plate using 20 microsatellite markers
in the HLA region (Table 2); these markers were described previously [6]. Seven
newly designed markers in the HLA region were used for genotyping all 384
samples in experiment 2 (Table 2). Primer sequences were selected from NCBI
sequence files (http://www.ncbi.nlm.nih.gov/genomes/sts).
Thermal cycling was performed on a Primus Multiblock HT PCR system (MWGBiotech, Ebersberg, Germany). PCR was started with incubation of 5 min at 95 °C,
26
Chapter 2
Table 2: Sequences of microsatellite markers
Expt.
1
2
Marker
6SL001
6SL002
DNRNGCA
RING3CA
D6S2658
6BO01
Tap1
G511525
D6S1666
D6S2670
D6S273
TNFα
MICA
D6S2673
D6S2678
D6S2694
D6S2699
D6S2700
D6S265
D6S2707
HL002
D6S2701
D6S2702
HL003
D6S2704
D6S2705
D6S510
Forward primer (5' → 3')
CCTCACCCGATACATAGACATAGG
CTCTCGCTACTGTGGTACATGC
AGGAATCTAGTGCTCTCTCC
TGCTTATAGGGAGACTACCG
AGAGAATGGATGCTGCATGAGG
AGGGAATTCGGAACTCATTTTT
AGAACCAGACAGGTTTCTCCTG
GGTAAAATTCCTGACTGGCC
CTTGAGGACTGAGTCTGAGTTGG
CCACCCACTTCCTCCACTAGAATC
GCAACTTTTCTGTCAATCCA
CCTCTAGATTTCATCCAGCCACA
GCCTTTTTTTCAGGGAAAGTGC
TTCTGCGTTTTCAGCCTGCTAG
TTGCAGTGAGCCAAGATCGC
TCTCTTTCCCAGTGTCCTTCTAAC
CGACTCCACCTATGACGGACATAC
CAGTTTCGCAACCTGTTTGCC
ACGTTCGTACCCATTAACCT
CAGTTTCGCAACCTGTTTGCC
TACCAGGTTGTAAGGCTCAACAT
GAGGTCTGTGGTCATAACTTTGG
ATAAAATCCAGGTCATGGTGGA
TTGAAAAACAGGTCATTTTTAGGTT
CCTTCTCTCCCCAAAGATAAACA
GCCTTCAGGACATGTTTGTGTGTA
AATGTTCCTGCTTTCATTTCTTT
Reverse primer (5' → 3')
AGAAATACCGAAATAAGGCCTCC
CAAACTGTAAGTCATGACCATGC
CTCTAGCAAAAGGAAGAGCC
GATGGGAAGTTTCCAGAGTG
TGTATAACCCGAAAGTCCAGCTCTC
GTAAACTGGGCTGAGATGTACCA
GGACAATATTTTGCTCCTGAGGTA
GACAGCTCTTCTTAACCTGC
GAATCCAGCATTTTGGAGTTGT
GTGAATTGTGACTGTGCCAGTACAC
GACCAAACTTCAAATTTTCGG
GCCTCTCTCCCCTGCAACACACA
CCTTACCATCTCCAGAAACTGC
GAACCACTCTTCGTACCACAGTCTC
CCCCACAAAAAACCCCTGTTTATC
GCAATACAGCAAGACCCTGTC
CCTCTTCTCAGCTCTTCCATCTCAC
GCATCAGCAGTCATTAGGGAAATGC
ATCGAGGTAAACAGCAGAAA
GCATCAGCAGTCATTAGGGAAATGC
GGCTGAGATGAGAGAATCACTTC
TGTGGTTTCATTTCCTTCTAGTCA
GGCCTAAATGCTTCCTTGGATA
GGGCAACAAGATCAAAACTCTG
GTAATTTTTGCCACTCTGGAGGA
TTCAACTCTTTTAGCTGTTTTGG
GTCAAAACTGCAATGGGCTACTA
Table 3: Factorial design for three factors each at two levels
Reaction
Factor 1
Factor 2
Factor 3
mixture
1
1
1
1
2
1
1
2
3
1
2
1
4
1
2
2
5
2
1
1
6
2
1
2
7
2
2
1
8
2
2
2
In the table 1 and 2 represent -10% and +10% of the standard factor condition; all possible combinations result in
8 reaction mixes.
followed by 30 or 34 cycles (−2 and +2 of standard used cycles) for amplification
with the following thermal profile: 94 °C for 30 s, primer-specific annealing
temperature (Ta) (±2 °C of optimal temperature) for 30 s, 72 °C for 1 min and the end
of the last cycle was 72 °C for 5 min. The PCR conditions were tested as outlined in
Optimizing PCR conditions
this paper. Subsequent to the PCR, products were pooled into predefined panels,
according to allele length and fluorescent label, by use of a Biomek 2000 pipetting
robot (Beckman Coulter, Allendale, NJ, U.S.A.). A sample (2.3 μl) of the pooled
products was mixed with 2.5 μl of MilliQ water and 0.2 μl of ET-400R size standard
(Amersham Biosciences). The products were visualized by separating the samples
on a MegaBACE 1000 capillary sequencer (Amersham Biosciences) according to the
manufacturer’s instructions. Genotyping steps other than the PCR were kept
constant during the experiments.
Experimental design
For experiment 1, we used a factorial design of eight factors at two levels, which
resulted in 256 different conditions. Table 3 shows an example of a 23 design. Six of
the factors in the PCR varied in concentration: PCR buffer, MgCl2, dNTPs, primers,
DNA and Taq. The two concentration levels applied for these variables were −10%
and +10% of standard concentrations (Table 1). The other two factors varied were
the annealing temperature (Ta) with two levels (Ta=−2 and +2 °C) and the number of
PCR cycles (30 and 34 cycles). All possible combinations of the five factors in the
PCR reaction mixture, PCR buffer, MgCl2, dNTPs, primers and Taq, resulted in 32
different mixtures. These mixtures were systematically divided over two
concentration levels of the DNA samples in a 384-well plate. These 384-well plates
containing the DNA samples and the different mixtures were divided over the two
levels of PCR cycles and annealing temperatures. The experiment was performed on
384 samples for each of the 20 microsatellite markers, so the conditions were divided
over a total of 7680 samples; therefore each specific condition was performed 30
times.
For the second experiment, three factors with two levels were tested; these were the
significant factors resulting from experiment 1, except for the factor of buffer. The
buffer also contained MgCl2 and this could interact with the factor, additional
MgCl2. The factors MgCl2 and Taq varied with −10% and +10% of optimal
concentrations, and the Ta varied with −1 and +1 °C of optimal Ta (Table 1).
Experiment 2 was performed on 384 samples for seven microsatellite markers, and
the eight conditions were systematically divided over 2688 samples, so each specific
condition was performed 336 times. For both experiments, each DNA sample had a
fixed position in the plates.
Analysis
First-pass genotyping results were used for analysis. Lengths of the alleles were
visualized and analysed using Genetic Profiler version 2.0 (Amersham Biosciences).
The quantity of the PCR product produced was measured by use of the peak height
(relative fluorescence) of the short allele. The dataset was prepared using Excel 2000
27
28
Chapter 2
Table 4: Results of ANOVA (mean effects of experiment 1)
Factors
Type III Sum of
Squares
Degrees of
freedom
F-value
P-value
MgCl2
Ta
630
1
276.53
0
280
1
123
0
Taq
1070
1
471.55
0
Marker
5540
19
121.4
0
Buffer
750
1
329.91
0
dNTP
3.28
1
1.44
0.23
primer
141
1
61.78
0
DNA
5.23
1
2.3
0.129
60
0
Cycles
136
1
Errora
16000
7052
24700
7079
Corrected Totalb
Dependent variable: peak height of the shortest allele. R2 =0.350.
a The variation left unexplained after the model has been considered.
b The total variation in the dependent variable, corrected for the mean.
Table 5: Mean peak heights in experiment 1
Factor
MgCl2
Ta
Taq
Buffer
Variation
Mean peak
height
SEM
95% confidence interval
Lower
Upper
N
-10%
12639.2
264.2
12121.4
13157.0
3533
+10%
18602.3
264.2
18084.5
19120.2
3547
-2°C
17603.1
265.3
17083.1
18123.2
3515
+2°C
13638.4
263.0
13122.8
14154.0
3565
-10%
11699.9
261.3
11187.7
12212.1
3580
+10%
19541.6
267.2
19017.9
20065.3
3500
-10%
18980.4
260.1
18470.6
19490.2
3475
+10%
12452.6
252.4
11957.8
12947.3
3605
(Microsoft Office). The genotyping success was the percentage of alleles for all
markers that could be determined. The conditions producing the maximum PCR
product was analysed by comparing the peak heights of the different conditions by
the use of ANOVA, using SPSS version 11.0 (SPSS, Chicago, IL, U.S.A.). The eight
factors were analysed, including interactions between these factors for a significant
effect on the quantity of PCR product. The presence or absence of data (representing
total failure of PCR) was analysed as a dependent variable as well. Homozygosity of
the genotype was introduced as covariate, because homozygous peaks are expected
to have a higher signal than heterozygous peaks.
Optimizing PCR conditions
29
Table 6: Results ANOVA (mean effects experiment 2)
Type III Sum
of Squares
Degrees of
freedom
F-value
P-value
MgCl2
1601
1
125.88
.000
Ta
11020
1
866.28
.000
Taq
239.54
1
18.83
.000
Factors
Marker
9097
6
119.21
.000
120.55
1
9.48
.002
MgCl2*Ta
2.41
1
0.19
.663
MgCl2*marker
5653
6
74.07
.000
Taq* Ta
30.87
1
2.43
.119
Taq*marker
2318
6
30.37
.000
Ta*marker
6687
6
87.62
.000
Taq* Ta*marker
1122
6
14.70
.000
17.45
.000
MgCl2*Taq
MgCl2*Taq*Ta*marker
4218
19
Errora
28370
2231
71550
2286
Corrected Totalb
Dependent variable: peak height of shortest allele, R2= 0.603. Asterisks indicate the interaction between the
factors (interaction term). If the P-value is significant, it means that the factors interact with each other.
a The variation left unexplained after the model has been considered.
Results
Experiment 1
The mean peak heights of the different markers for experiment 1 were between 5600
and 42 000. Genotyping success was 92.7%, and the effects of homozygosity on
amplitude were negligible. The results of ANOVA are shown in Table 4. Factors that
did not influence the peak heights (quantity of PCR product) were the
concentrations of DNA, dNTPs, primers and the number of PCR cycles. The factors
MgCl2, buffer, Taq, Ta and the marker did significantly affect the signal height (for all
the mentioned effects P<0.001, and with a high F value). In general, an increased
concentration of MgCl2 and Taq, a decreased concentration of buffer and a lower Ta
value gave higher signals (Table 5). An increase of Taq concentration by 20%
resulted in signal heights that were almost twice as high. It was not possible to
identify the best combination of conditions, since the factor MgCl2 was confounded
with other factors, due to an error in the plate setup. In experiment 2, we therefore
analysed the best concentrations and temperatures for the interactions between the
factors having the strongest main effects as seen in experiment 1.
Experiment 2
For experiment 2, the mean peak heights were between 24 000 and 45 000, and
genotyping success was 97.8%. The overall variation in amplitude was explained for
30
Chapter 2
60% (R2 = 0.603) by all the experimental variations combined, which means that the
variation in amplitude is largely controlled by the experimental conditions. The
effects of homozygosity on amplitude were negligible. The results of ANOVA are
shown in Table 6; the main factors of experiment 1 (MgCl2, Taq and Ta) remained
important (for all the mentioned effects P<0.001) and the marker also influenced the
signal height (P<0.001). The significant interaction between the factors MgCl2, Taq, Ta
and the marker (P<0.001) indicates that more than an additive signal is present at
different combinations of conditions for each marker. The combination of conditions
that produced, overall, the highest signals was +10% MgCl2, +10% Taq and a lower Ta
value, which is consistent with the main effects. Stable conditions are defined as
minimally variable due to change in one of the conditions. Table 7 shows stable
conditions for three randomly chosen markers. For marker HL002, the peak heights
were not influenced by the experimental variations. The most stable condition for
the markers D6S2701 and HL003 occurred at the condition +10% MgCl2, +10% Taq
and Ta= −2 °C. When two factors (MgCl2 and Taq) were together reduced, average
signal height decreased from 43 035 to 26 072. Conditions for the markers D6S2704,
D6S2705 and D6S510 were in particular stable by using a lower Ta. Table 7 also
shows results for the marker D6S2702; this marker was most stable under the
condition: −10% MgCl2, −10% Taq and Ta=−2 °C or +2 °C. When the concentrations of
both MgCl2 and Taq were increased, peak height decreased from 40 682 to 28 011.
Discussion
With this experiment, we have shown that a factorial experiment can be used
without affecting the genotyping success, since success rates were still 92.7 and
97.8%, for experiments 1 and 2 respectively. Experiment 1 showed that, in general,
an increased concentration of MgCl2 and Taq, a decreased concentration of buffer
and a lower Ta value resulted in an increased quantity of PCR product. In
experiment 2, these significant factors were varied, except for the factor of buffer,
since the factor of buffer containing MgCl2 interacts with the factor of additional
MgCl2. An increase of the MgCl2 concentration would lead to decrease of MgCl2 in
the buffer, and the effect shown by the increase of MgCl2 is also caused by increase
of MgCl2 in the buffer. The results of the main factors tested in experiment 2 were
the same as those for experiment 1, so an increase of MgCl2 did indeed increase the
amount of PCR product. With too little Mg2+, the polymerase will have poor activity,
but, with too much Mg2+ [2] and a low Ta value, non-specific amplification could
become a problem. However, production of non-specific amplification products was
not obvious or present in too low concentrations to affect the results. The condition
is not universal for all markers, because the effect of the microsatellite marker is
high in both the experiments. This could be explained by marker-specific effects,
Optimizing PCR conditions
31
Table 7: Most stable peak heights for the markers D6S2701, D6S2705 and D6S2702
Marker
D6S2701
Ta
−2 °C
Taq
−10%
+10%
+2 °C
−10%
+10%
D6S2705
−2 °C
−10%
+10%
+2 °C
−10%
+10%
D62702
−2 °C
−10%
+10%
+2 °C
−10%
+10%
MgCl2
−10%
+10%
−10%
+10%
−10%
+10%
−10%
+10%
−10%
+10%
−10%
+10%
−10%
+10%
−10%
+10%
−10%
+10%
−10%
+10%
−10%
+10%
−10%
+10%
Mean peak height
26 072
42 663
40 402
43 035*
16 833
41 223
21 369
39 539
53 231*
53 583*
51 686*
56 034*
30 762
8 800
29 730
33 335
40 682*
39 894
35 926
28 011
40 366*
36 707
36 283
29 919
since the efficiency of amplification is also influenced by the specific sequence of the
target site and primers [8]. Nevertheless, the condition producing the highest
amount of PCR product over all markers on average might be a good starting point
for optimizing new microsatellite markers. Variations in the concentrations of DNA,
dNTPs and primers and in the number of PCR cycles did not affect the quantity of
PCR product. Therefore it is likely that the concentrations of these factors were
higher than necessary; these concentrations and the number of cycles might be
reduced to achieve a possibly more optimal condition and to lower the costs. This
experiment shows that our current working conditions were not optimal.
In the second experiment, we showed that there was a significant interaction
between the factors MgCl2, Taq and Ta. The combination of the main effects gave the
same result as indicated by the interaction. The condition with the highest quantity
of PCR product was not always the best operational condition, since the second
experiment demonstrated the presence of a robust set of conditions with good
signals.
The use of a stable condition is recommended because, under unstable conditions,
only one small variation in MgCl2 or Taq concentration did strongly reduce peak
heights. Variations in Ta also had a strong effect on the quantity of PCR product, but
generally this factor is well controlled during the PCR process.
32
Chapter 2
We recommend choosing a stable condition as the working point, accepting a
marginally lower signal rather than one with maximum signal. Although,
apparently, a single robust and optimal set of amplification conditions for all
markers does not exist, a robust condition for each marker with little sacrifice in
signal was easily identified. For high-throughput genotyping, it is essential to check
continuously whether operating conditions are maintained, because small
differences due to instrument drift or dilution errors may result in quite large
differences in signal strength and genotyping success. A factorial design of the type
that we used is very sensitive and can be routinely applied and operated with
smaller experimental changes than we applied, in the order of a few percentage
change, to verify that conditions are still optimal. The high proportion of explained
variance (60%) with regard to signal strength indicates that we now have control
over a large part of the quality-defining operating conditions, given the substantial
size of random uncontrolled variations. By use of this design, many highthroughput techniques, using different factors, can be checked for their optimality of
working conditions. With current robotics, the necessary systematic variations can
be produced routinely without much effort. This implies that important process
information can be obtained at little cost.
Acknowledgements
We thank Dr G. van der Steege (Department of Medical Biology, University Medical Center Groningen) and
technicians for marker design and DNA isolation respectively. We also thank Dr E. Vellenga, Dr G. W. van
Imhoff (Department of Hematology) and Dr S. Poppema (Department of Pathology) of the University Medical
Centre Groningen, The Netherlands, for co-designing the Hodgkin study, which we used for our experiments.
This research work was supported by a grant from the Dutch Cancer Society (KWFNKB 99-1878) and by
Genizon Biosciences (Montreal, QC, Canada).
References:
[1]
Moretti T, Koons B, Budowle B. Enhancement of PCR amplification yield and specificity using AmpliTaq
Gold DNA polymerase. Biotechniques. 1998; 25: 716-22.
[2]
Cobb BD, Clarkson JM. A simple procedure for optimising the polymerase chain reaction (PCR) using
modified Taguchi methods. Nucleic Acids Res. 1994; 22: 3801-5.
[3]
Spijker GT, Bruinenberg M, te Meerman GJ. Efficiency control in large-scale genotyping using analysis of
variance. Appl Biochem Biotechnol. 2005; 120: 29-36.
[4]
Siouffi AM, Phan-Tan-Luu R. Optimization methods in chromatography and capillary electrophoresis. J
Chromatogr. 2001; A892: 75–106.
[5]
Benčina M. Optimisation of multiple PCR using a combination of full factorial design and threedimensional simplex optimisation method. Biotechnol Lett. 2002; 24: 489–95.
[6]
Diepstra A, Niens M, Vellenga E, van Imhoff GW, et al. Association with HLA class I in Epstein-Barrvirus-positive and with HLA class III in Epstein-Barr-virus-negative Hodgkin's lymphoma. Lancet. 2005;
365: 2216-24.
[7]
Sambrook J, Fritsch EF and Maniatis T. Molecular cloning: a laboratory anual. Cold Spring Harbor
Laboratory Press, Plainview, NY. 1989.
[8]
Rochelle PA, De Leon R, Stewart MH, Wolfe RL. Comparison of primers and optimization of PCR
conditions for detection of Cryptosporidium parvum and Giardia lamblia in water. Appl Environ
Microbiol. 1997; 63: 106-14.