Performance of Whole-Genome Amplified DNA Isolated from Serum and Plasma on High-Density Single Nucleotide Polymorphism Arrays

Abstract

Defining genetic variation associated with complex human diseases requires standards based on high-quality DNA from well-characterized patients. With the development of robust technologies for whole-genome amplification, sample repositories such as serum banks now represent a potentially valuable source of DNA for both genomic studies and clinical diagnostics. We assessed the performance of whole-genome amplified DNA (wgaDNA) derived from stored serum/plasma on high-density single nucleotide polymorphism arrays. Neither storage time nor usage history affected either DNA extraction or whole-genome amplification yields; however, samples that were thawed and refrozen showed significantly lower call rates (73.9 ± 7.8%) than samples that were never thawed (92.0 ± 3.3%) (P < 0.001). Genotype call rates did not differ significantly (P = 0.13) between wgaDNA from never-thawed serum/plasma (92.9 ± 2.6%) and genomic DNA (97.5 ± 0.3%) isolated from whole blood. Approximately 400,000 genotypes were consistent between wgaDNA and genomic DNA, but the overall discordance rate of 4.4 ± 3.8% reflected an average of 11,110 ± 9502 genotyping errors per sample. No distinct patterns of chromosomal clustering were observed for single nucleotide polymorphisms showing discordant genotypes or homozygote conversion. Because the effects of genotyping errors on whole-genome studies are not well defined, we recommend caution when applying wgaDNA from serum/plasma to high-density single nucleotide polymorphism arrays in addition to the use of stringent quality control requirements for the resulting genotype data.

Recent technical advances have produced platforms to assay hundreds of thousands of genetic polymorphisms across the human genome. High-density single nucleotide polymorphism (SNP) genotyping panels with automated allele calling are rapidly becoming powerful tools for identifying new genes that contribute to common diseases,^1,2 detecting copy number variations,³ and for novel applications in forensics⁴ and anthropology.^5,6 Genome-wide SNP typing requires sufficient amounts of high-quality DNA, which is usually obtained from whole blood, cell lines, and surgical biopsies. Unfortunately, the amount of DNA is often limited in many repositories where stored serum, buccal swabs, or dried blood spots are the only sources of DNA available.

Whole-genome amplification is an important technology for genomic and molecular diagnostic studies that permits analysis of samples containing only minute amounts of DNA.⁷ Early protocols for whole-genome amplification using TaqDNA polymerase with degenerate primers^8,9 were hampered by amplification artifacts, allelic dropout, and localized enrichment leading to incomplete genomic coverage. New amplification systems based on multiple displacement amplification and the highly processive ϕ29 DNA polymerase¹⁰ provide highly uniform representation across the genome regardless of DNA sequence or secondary structure, high-fidelity amplification with low error rates of only 1 in 10⁶ to 10⁷ nucleotides, and high yields of amplified DNA, as much as 10⁷-fold amplification.¹¹,¹²,¹³ Resulting whole-genome amplified DNA (wgaDNA) can be directly used in microarray-based methods for genomic analysis, such as SNP genotyping¹⁴ and comparative genomic hybridization.¹⁵

Previous studies of relatively small numbers of polymorphisms have shown high concordance between genomic DNA (gDNA) and wgaDNA from a variety of sources,^11,16,¹⁷,¹⁸ but to our knowledge, no studies have evaluated performance of wgaDNA on very large numbers of SNPs using high-density arrays. To assess the potential utility of serum/plasma DNA in molecular epidemiology studies and diagnostic applications, we examined performance of wgaDNA on Affymetrix 500K arrays containing 500,568 SNPs. Our specific objectives were to assess call rates of wgaDNA from samples that had been stored for extended periods of time (up to 20 years), to determine rates of concordance in genotyping data between gDNA and wgaDNA, and to characterize the nature and genomic distribution of discordant genotypes. Information on potential uses and limitations of wgaDNA obtained from stored serum/plasma samples will be important for guiding future uses of wgaDNA in large-scale genomic studies.

Materials and Methods

Sample Collection

All research was conducted under protocols approved by the Windber Medical Center and University of Pittsburgh Institutional Review Boards. Fresh serum and plasma samples were obtained from subjects who voluntarily agreed to participate in this study and gave written informed consent. Peripheral blood was collected in serum separator or 10-ml heparinized BD vacutainers (Thermo Fisher Scientific, Waltham, MA) and placed at room temperature for 30 minutes (serum) or on wet ice (plasma). All vacutainers were centrifuged at 1300 × g for 10 minutes at 4°C, then divided into 500-μl aliquots and stored at −80°C. Stored serum/plasma samples that were originally collected 3, 10, or 20 years ago were obtained from the sample archives of the Integrative Cardiac and Metabolic Health Program at Windber Research Institute and the Epidemiology of Diabetes Complications Study at the Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh.

DNA Isolation from Serum and Plasma

Before DNA isolation, samples were removed from −80°C, allowed to thaw on ice for 30 minutes, then mixed thoroughly. To simulate sample usage over time, which is often characteristic of banked serum/plasma samples, aliquots of two recently collected serum samples were thawed and refrozen five times. DNA was extracted from 200 μl of serum or plasma using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany). Samples were treated with Qiagen protease, then buffer AL (lysis solution) at 56°C for 10 minutes using a TropiCooler (Boekel Scientific, Feasterville, PA). DNA was isolated in QIAamp spin columns, then eluted, dried, and reconstituted in a final volume of 100 μl of AE buffer. Initial DNA concentrations and purity were determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE). Visual inspection of representative serum/plasma DNA aliquots on agarose gels showed a prominent band >10 kb in size with minimal detectable degradation.

For all samples where whole blood was available, gDNA was isolated from peripheral blood mononuclear cells with the Puregene DNA purification kit (Gentra Systems, Minneapolis, MN) according to the manufacturer's protocol. Concentration and purity of the gDNA was determined as above on the NanoDrop ND-1000 spectrophotometer.

Whole-Genome Amplification of Serum/Plasma DNA

Serum or plasma DNA aliquots (100 μl) were concentrated to a final volume of 10 μl at 45°C in an Eppendorf Vacufuge concentrator (Brinkman Instruments, Westbury, NY).¹⁹ Whole-genome amplification was then performed on 2.5 μl of concentrated serum/plasma DNA subjected only to alkaline denaturation using the REPLI-g whole-genome amplification kit (Qiagen). For samples that had been thawed and refrozen, a second independent whole-genome amplification was performed using template DNA that had been subjected to both alkaline and heat (3 minutes at 95°C) denaturation to determine whether multiple whole-genome amplifications could be pooled to increase SNP call rates on 500K arrays. All amplifications were incubated for 16 hours at 30°C in the TropiCooler, followed by inactivation of the REPLI-g DNA polymerase at 65°C for 3 minutes. Amplified serum/plasma DNA was then purified using the ethanol precipitation purification protocol in the GenomiPhi DNA amplification kit (GE HealthCare, Buckinghamshire, UK). Purified wgaDNA was re-suspended in 30 μl of TE buffer, assessed on the NanoDrop ND-1000, and then stored at −20°C until used for genome-wide SNP analysis.

SNP Analysis

Purified wgaDNA samples (diluted to 50 ng/μl) from serum/plasma and gDNA from whole blood were applied to Affymetrix GeneChip Human Mapping 500K arrays following the manufacturer's protocol. The DNA was cleaved with either NspI or StyI restriction enzymes, ligated to specific linkers, and amplified with the Titanium DNA amplification kit (Clontech, Mountain View, CA). Polymerase chain reaction products were purified using the Clontech DNA amplification clean-up kit, diluted to 90 μg in 45 μl of Clontech RB buffer, fragmented, and labeled with biotinylated GeneChip DNA labeling reagent. Samples were then hybridized to the Affymetrix GeneChip Human Mapping 500K arrays. Following hybridization, arrays were stained and scanned on a GeneChip scanner 3000. To assess the effects of a second whole-genome amplification on call rates for thawed-refrozen samples, independent reactions were pooled before SNP analysis.

Data Analysis

For all analyses, P < 0.05 was used to determine statistical significance. Genotypes (AA, AB, BB, or no call) were determined using the dynamic model mapping algorithm in the Affymetrix GeneChip genotyping analysis software (GTYPE 4.0) package²⁰ at a 0.33 confidence score. For wgaDNA, SNP call rates were compared between NspI and StyI arrays across all samples using a paired t-test. As no significant difference (P = 0.10) in call rates was detected, NspI and StyI results were combined in subsequent analyses. We used a Wilcoxon signed-rank test, which accounts for a non-normal distribution of differences between pairs, to compare genotype call rates for gDNA versus wgaDNA and single versus multiple whole-genome amplifications from the same participant. A Pearson correlation was used to examine variability in call rates as a function of storage time in never-thawed samples and Student's t-test was used to assess call rates for never-thawed serum/plasma aliquots versus samples that had been thawed and refrozen at least once over time.

Genotypes were imported into Access 2003 (Microsoft, Redmond, WA) and queried to determine levels and patterns of concordance between gDNA and wgaDNA. Spotfire DecisionSite (TIBCO Software, Palo Alto, CA) was used to examine the chromosomal distribution of SNPs that were not scored consistently between gDNA and wgaDNA. Approximate chromosomal locations for SNPs showing discordant genotypes or homozygote conversion were obtained by NetAffx (available from Affymetrix) batch queries.

Results

DNA Extractions and Whole-Genome Amplifications

DNA was extracted from 200-μl aliquots of plasma or serum collected over a period of 20 years from 12 individuals (Figure 1). Average DNA yields (±SD) were 366 ± 27 ng for recently collected samples, 454 ± 77 ng for samples that had been stored for 3 years, 412 ± 14 ng for 10-year samples, and 322 ± 85 ng for 20-year-old aliquots. Yields were not correlated with storage time of the sample (r² = 0.18, P = 0.17). The OD 260/280 ratio, a measure of DNA purity with respect to protein contamination, ranged from ∼1.3 to ∼1.7 (average 1.4 ± 0.1).

Figure 1.

DNA extraction (A) and whole-genome amplification (B) yields from 200-μl aliquots of serum or plasma collected over a 20-year period from 12 individuals.

When aliquots of serum/plasma DNA (∼100 ng) were subjected to whole-genome amplification, with the exception of one sample that may have been compromised during storage and thus failed to amplify, all reactions generated more than 30 μg of wgaDNA (Figure 1). Average yield of the amplified products (±SD) from alkaline-denatured template was 37.5 ± 3.8 μg; average yield in the subset of thawed-refrozen samples with a second amplification from template that had been subjected to alkaline and heat denaturation was 32.6 ± 9.1 μg. All corresponding OD 260/280 ratios were consistent at 1.9 ± 0.02. Multiple freeze-thaw cycles did not reduce the efficiency of DNA extraction from serum (yield ≥400 ng of DNA) or the whole-genome amplification process (amplifications generated >40 μg wgaDNA).

SNP Genotyping Quality Control

The mapping algorithm report in GTYPE contains several parameters that are useful for assessing quality of 500K SNP genotyping results. Call rates for the reference genomic DNA 103 (96.5% for NspI, 97.8% for StyI) indicated good overall performance of the assays. The modified partitioning around medoids (MPAM) algorithm, developed to make genotype determinations on the GeneChip Human Mapping 10K arrays, genotypes a subset of ∼8000 SNPs on the 500K arrays. For high-quality genomic DNA, the MPAM call rate (MCR) should be ≥94% and the MPAM detection rate (MDR) should be ∼99%. In our gDNA samples, all parameters were within the manufacturer's recommended specifications: MCRs were ≥94.2% and MDRs were ≥98.7% (data not shown). Although comparable QC parameters have not been developed for wgaDNA, our wgaDNA samples showed MCR values ranging from ∼20% to 93% and MDR values of ∼31% to 99% (Table 1). Both MCR and MDR values were significantly higher (P < 0.001) in never-thawed serum/plasma samples compared to previously thawed and refrozen samples.

Table 1.

Parameters for Assessing Quality of 500K Data Generated from wgaDNA

			QC parameter
			MCR		MDR
Sample	Storage time	Date of collection	NspI	StyI	NspI	StyI
109022 serum	Recent	11/02/2006	73.6	71.9	92.3	91.3
109023 serum	Recent	04/17/2007	76.8	79.2	94.6	96.0
109025 serum	Recent	04/17/2007	72.0	75.2	91.2	92.7
0270 plasma	∼3 years	01/12/2004	90.2	82.5	98.4	95.3
0271 plasma	∼3 years	01/12/2004	93.1	92.5	98.8	98.2
0905 serum	∼10 years	02/08/1997	63.5	68.6	88.5	93.0
8773 serum	∼10 years	03/05/1997	79.3	81.7	96.5	97.4
0625 serum	∼20 years	09/17/1988	85.9	92.6	96.7	98.8
0531 plasma	∼20 years–TR	04/04/1987	19.9	38.1	31.4	49.2
1166 plasma	∼20 years–TR	04/01/1987	53.2	58.7	66.5	70.6
109023 serum	Recent–5X	04/17/2007	34.6	38.1	44.6	49.2
109025 serum	Recent–5X	04/17/2007	57.3	58.7	74.2	70.6

From the dynamic mapping algorithm report in GTYPE 4.0. ∼20 years—TR, archived samples that were thawed and refrozen for other purposes at least once over time; recent–5X, samples experimentally thawed and refrozen five times to simulate heavy sample usage. NspI arrays contain 262,264 SNPs, StyI arrays contain 238,304 SNPs.

Genotype Call Rates

Call rates for all gDNA samples were >97%, well above the Affymetrix recommended minimum call rate of 93% (Table 2). In wgaDNA samples, call rates showed more variability, ranging from ∼63% in a 20-year-old sample that had been thawed and refrozen at least once to ∼97% in a 20-year-old never-thawed serum sample. The percentage of genotype calls did not differ significantly between gDNA and wgaDNA (P = 0.13) for samples that were recently collected or stored for approximately 3 years (Table 2). For wgaDNA from never-thawed serum/plasma samples collected at various times over a 20-year period, no significant correlation was observed between call rates and age of the sample (r² = 0.08, P = 0.51), indicating that call rates did not diminish with increasing storage time. However, samples that had been thawed and refrozen, either experimentally or due to usage over time, showed significantly lower call rates (P < 0.001) than never-thawed samples. Pooling of two independent whole-genome amplifications with different initial denaturation conditions improved call rates in the thawed-refrozen samples by 2.3 ± 3.5% (P < 0.05).

Table 2.

SNP Call Rates (%) on 500K Arrays from gDNA and wgaDNA

			DNA
			gDNA		wgaDNA
Sample	Storage time	Date of collection	NspI	StyI	NspI	StyI
109022 serum	Recent	11/02/2006	—	—	89.1	89.0
109023 serum	Recent	04/17/2007	97.4	97.3	91.4	93.0
109025 serum	Recent	04/17/2007	97.3	97.7	89.0	90.9
0270 plasma	∼3 years	01/12/2004	97.9	97.1	95.4	91.4
0271 plasma	∼3 years	01/12/2004	97.8	97.8	96.1	95.8
0905 serum	∼10 years	02/08/1997			85.1	88.2
8773 serum	∼10 years	03/05/1997			92.6	93.6
0625 serum	∼20 years	09/17/1988			94.1	96.8
0531 plasma	∼20 years–TR	04/04/1987			62.9	65.0
1166 plasma	∼20 years–TR	04/01/1987			79.2	81.6
109023 serum	Recent–5X	04/17/2007			69.7	70.5
109025 serum	Recent–5X	04/17/2007			80.9	81.6

∼20 years–TR, archived samples that were thawed and refrozen for other purposes at least once over time; recent–5X, samples experimentally thawed and refrozen five times to simulate heavy sample usage. NspI arrays contain 262,264 SNPs, StyI arrays contain 238,304 SNPs.

Concordance Between wgaDNA and gDNA

To evaluate whether wgaDNA gives robust genotype data on high-density SNP arrays, we determined concordance between wgaDNA and gDNA considering only SNPs that gave genotype calls in both amplified and unamplified samples. For all individuals where gDNA was available for study, approximately 400,000 or more SNPs (∼80%+ of all SNPs on the 500K arrays) yielded concordant genotype calls between gDNA and wgaDNA (Table 3). In all cases, discordance rates were <10%; average discordance (±SD) was 4.4 ± 3.8%. Heterozygote dropout in wgaDNA accounted for the majority of discordant genotypes, ranging from 60% to 91% of all discrepancies (average 78.5 ± 14.1%; see Table 4). All other differences comprised only a minor component of the overall discordance in genotype calls between gDNA and wgaDNA.

Table 3.

Concordance Rates for SNP Genotyping on 500K Arrays Between gDNA and wgaDNA

Sample	No. unscored SNPs in gDNA	No. unscored SNPs in wgaDNA	No. concordant genotypes	% concordance	No. discordant genotypes	% discordance
109023–NspI	6862	22,537	219,491	83.7	17,376	6.6
109023–StyI	6365	16,652	203,331	85.3	15,298	6.4
109025–NspI	7027	28,945	205,991	78.5	24,663	9.4
109025–StyI	5468	21,690	193,427	81.2	21,055	8.8
0270–NspI	3957	10,610	244,516	93.2	3181	1.2
0270–StyI	6912	20,549	212,074	89.0	4081	1.7
0271–NspI	5714	10,276	247,696	94.4	1656	0.6
0271–StyI	5156	10,086	224,844	94.4	1572	0.7

NspI arrays contain 262,264 SNPs; StyI arrays contain 238,304 SNPs. The number of concordant and discordant genotypes includes only those SNPs that gave genotype calls in both amplified (wgaDNA) and unamplified (gDNA) samples, thus % concordance + % discordance does not equal 100%.

Table 4.

Nature of Discordant Genotypes from 500K Arrays Between gDNA and wgaDNA

Nature of discordance		109023			109025			0270			0271
gDNA genotype	wgaDNA genotype	NspI	StyI	Total % of discordant genotypes	NspI	StyI	Total % of discordant genotypes	NspI	StyI	Total % of discordant genotypes	NspI	StyI	Total % of discordant genotypes
AA	BB	64	43	0.3	214	134	0.8	1	26	0.4	2	1	0.1
BB	AA	72	53	0.4	257	195	1.0	5	22	0.4	3	2	0.1
AA	AB	700	607	4.0	1389	990	5.2	375	447	11.3	278	278	17.2
BB	AB	735	558	4.0	1508	1068	5.6	452	445	12.4	361	371	22.7
AB	AA	8302	7365	47.9	11,330	9773	46.2	1279	1710	41.1	529	525	32.7
AB	BB	7503	6672	43.4	9965	8895	41.2	1069	1431	34.4	483	395	27.2
Total		17,376	15,298		24,663	21,055		3181	4081		1656	1572

NspI arrays contain 262,264 SNPs; StyI arrays contain 238,304 SNPs. Total % of discordant genotypes includes discordant genotypes from both NspI and StyI chips for each sample.

Analysis of the genomic distribution of SNPs yielding discordant genotypes showed no distinct patterns of chromosomal clustering. When all discordant genotypes (including allelic dropout) across multiple patients were combined, the distribution appeared relatively uniform throughout the genome (Figure 2). Likewise, SNPs showing homozygote conversion, a pattern of discordance involving homozygous genotypes for the alternate allele (eg, AA in gDNA, BB in wgaDNA), appeared to be evenly distributed (Figure 3). Complete data have been deposited to the NCBI dbSNP database (http://www.ncbi.nlm.nih.gov/projects/SNP/snp_viewTable.cgi?handle=WINDBER_RESEARCH).

Figure 2.

Genomic distribution of SNPs yielding discordant genotypes between wgaDNA derived from stored serum/plasma and gDNA isolated from peripheral blood on NspI (A) and StyI (B) arrays. The centromeres are indicated by gray bars.

Figure 3.

Genomic distribution of SNPs showing homozygote conversion between wgaDNA derived from stored serum/plasma and gDNA isolated from peripheral blood on NspI (A) and StyI (B) arrays. The centromeres are indicated by gray bars.

Discussion

Major efforts to define genetic variation associated with complex human diseases require high-quality DNA from well-characterized patients. Many studies with abundant long-term demographic and risk factor information, however, have only limited amounts of DNA available for genomic studies. Serum banks²¹ for example, often contain tens of thousands of samples and traditionally have been used for plasma-based screening in diverse epidemiological surveys.^22,23 With the development of robust technologies for whole-genome amplification, these repositories now represent a potentially valuable source of DNA for genomic studies and molecular diagnostic uses.

Whole-genome amplification has been performed successfully on DNA recovered from several different body fluids including saliva,²⁴ urine,¹⁸ amniotic fluid,²⁵ and plasma,¹⁹ and in general, the resulting SNP genotype data from wgaDNA were >97% concordant with SNP data from gDNA for small numbers of SNPs. Studies using whole-genome amplification to rescue DNA from plasma samples stored under optimal conditions have observed genotyping success rates of >93% and error rates of <1% for eight SNPs.¹⁹ Despite the robust SNP genotyping results attained with high-quality samples, little information exists regarding the effects of suboptimal long-term storage and thawing and refreezing of serum/plasma samples on the genotyping performance of wgaDNA using very high density SNP arrays.

Collections of archived biological samples normally experience a gradual deterioration in DNA quality over time, which may be accelerated by repeated thawing and refreezing as samples are used to assess other biomarkers. Overall call rates and accuracy of genotyping results in studies that rely on whole-genome amplification can be highly variable,²⁶ but generally correlate with the quality and/or quantity of the template DNA being used in the amplification reactions. Poor-quality DNA when used as a template for whole-genome amplification can reduce the efficiency of the reactions and increase the prevalence of artifacts such as loss of uniform genomic coverage and unbalanced amplification of alleles at specific sites (allelic dropout).²⁶,²⁷,²⁸,²⁹ Such genotyping errors decrease statistical power for detecting associations between genetic variants and complex disease phenotypes and can strongly affect individual identification in forensic cases.³⁰

To our knowledge, this study is the first to evaluate performance of wgaDNA derived from serum/plasma on large-scale SNP genotyping and examine the effects of sample storage/usage on genotyping performance. We observed that frozen serum/plasma collected in different laboratories over a 20-year period produced DNA yields within the range of plasma DNA concentrations previously obtained in other laboratories using similar protocols.^31,32 Whole-genome amplification of the resulting serum/plasma DNA generated ample amounts (>30 μg) of relatively pure (OD 260/280 ratio ∼1.9) wgaDNA, comparable to previous investigations using similar technology.¹⁹ When applied to Human Mapping 500K arrays however, the wgaDNA samples showed significant variability in SNP call rates and this variability appeared to be affected by long-term storage conditions. In our study of a limited number of samples, overall call rates for wgaDNA from never-thawed serum/plasma did not differ significantly from call rates derived from high-quality gDNA and did not diminish with increasing storage time. Conversely, samples that had been thawed and refrozen showed significant deterioration in call rates compared to never-thawed samples.

A high degree of concordance in SNP genotype calls has been observed between unamplified (genomic) DNA and wgaDNA when high-quality gDNA was used as the template,^28,33 but degraded or very low copy number templates usually reduce the yield of wgaDNA and result in a higher frequency of genotyping errors.^34,35 Thawing and refreezing may alter the structural integrity of extracted DNA, leading to artifacts in the wgaDNA. In our experience, high yields of wgaDNA can be routinely generated from thawed and refrozen serum/plasma samples, but these samples have significantly lower call rates than never thawed samples. Therefore, multiple freeze-thaw cycles do not necessarily reduce the efficiency of DNA extraction from serum/plasma or the whole-genome amplification process, but the resulting wgaDNA is not likely to be of sufficient quality for accurate large-scale SNP genotyping.

Never-thawed serum/plasma, even when stored for extended periods of time, can yield accurate genotype data for very small numbers of markers,³⁶ but large numbers of SNPs on high density arrays may produce unacceptably high error rates even with well-preserved samples. Our data suggest that wgaDNA from serum/plasma may be useful for genomic studies and molecular diagnostic applications using Affymetrix 500K arrays, but stringent quality requirements need to be established. Because the discordance rate is closely associated with the overall call rate, the selection criterion could be a minimum call rate of 93%, as recommended by the manufacturer for genomic DNA. Using this selection strategy, only four of the eight never-thawed samples, and none of the thawed/refrozen samples, would be deemed acceptable. The average discordance rate for acceptable samples with gDNA available (0270 and 0271, Table 3) was 1.1%, which was equivalent to an average of 5245 genotyping errors per sample compared to genomic DNA. Unfortunately, the effects of genotyping error rates on whole-genome studies have not been well defined. Further research is needed to determine the number of genotyping errors and the distribution of these errors across samples that would lead to erroneous conclusions such as false positive or false negative associations.

Critical requirements for reliable SNP genotyping from wgaDNA include balanced amplification of different regions (uniform genomic coverage) and balanced amplification of both alleles at each genetic locus (no allelic dropout).²⁷ Previous studies have observed that specific regions of the genome may not be replicated efficiently during whole-genome amplification due to local base composition (GC content) and the presence of repetitive elements.³³ However, these under-represented regions do not appear to be restricted to specific areas such as centromeres and telomeres. When we examined the genomic location of SNPs showing discordant genotypes between gDNA and wgaDNA and SNPs showing homozygote conversion across multiple patients, the distribution appeared relatively uniform throughout the genome, suggesting that no genomic regions were more prone to genotyping errors than other regions. Therefore, we believe that 500K SNP data generated from high-quality wgaDNA will be unlikely to show localized loss of information in genome-wide association studies.

In conclusion, whole-genome amplification is a promising technology for a growing number of research and diagnostic applications, but the amount and quality of input DNA may limit applicability in many cases.³⁷ Archived samples with storage and usage histories that promote degradation of high molecular weight DNA appear more likely to produce suboptimal genotyping performance following whole genome amplification.³⁸ In this study, wgaDNA from serum/plasma samples stored for up to 20 years, but not subjected to thawing and refreezing, produced the best SNP genotyping results on Affymetrix 500K arrays; however, samples that had been thawed and refrozen showed significant deterioration in call rates compared to never-thawed samples. Numerous studies have investigated ways to preserve low-concentration solutions of DNA that are prone to degradation³⁹ and to improve the genotyping performance of wgaDNA from limited/degraded templates.^{26,28,29,32,40},⁴¹,⁴² In our hands, an independent whole-genome amplification significantly increased call rates for poorly performing samples, but only by ∼2%. We recommend using caution when applying wgaDNA from serum/plasma to high-density SNP arrays because even small discordance rates may lead to an unacceptably high number of genotyping errors. Given the enormous potential of serum banks for genetic testing, forensics, and genome-wide association studies, there is a critical need to further optimize whole-genome amplification methods to improve fidelity when amplifying partially degraded templates.⁴³,⁴⁴,⁴⁵,⁴⁶