This story was originally published on April 5.
Density matters when it comes to chromosomal microarrays, at least when it concerns the arrays' ability to detect absence of heterozygosity in a patient's sample.
A new study published last week in Genetics in Medicine compared the ability of three commercial array platforms of varying density and content sold by Affymetrix, Agilent Technologies, and Oxford Gene Technology to detect copy-number variation and AOH.
Elevated AOH levels can indicate an increased risk for recessive disorders due to regions of the patient's genome being identical by descent. Overestimation of AOH due to false-positives can confound further investigation and interpretation, and can potentially lead to a misdiagnosis, they wrote.
In the paper, the authors reported that while all three chips reliably detected copy-number variation, mosaicism, and uniparental isodisomy, AOH detection "varied significantly." Additional experiments and analyses led the authors to conclude that Affy's CytoScan, the array containing the greatest SNP content of the trio, "is best for detecting AOH."
Corresponding author Katie Rudd, a cytogenetics director at Emory Genetic Laboratory in Atlanta, said that she and her colleagues embarked on the comparison while validating OGT's CytoSure ISCA UPD SNP array. The UK company launched the array two and a half years ago (BAN 9/10/2010). Each slide contains four, 180,000-oligonucleotide arrays, each consisting of comparative genomic hybridization probes for detecting changes in copy number, as well as 6,000 SNPs to enable the detection of whole-chromosome uniparental disomy. UPD occurs when a person receives two copies of a chromosome, or part of a chromosome, from one parent and no copies from the other parent, exposing the individual to a greater risk of inheriting pathogenic variants.
To validate the array, EGL included AOH samples from patients with known UPD or from patients whose parents were known to be consanguineous. In both of these cases, the OGT array "detected long stretches of AOH as expected," Rudd said.
However, during the validation, EGL encountered some "surprising AOH results" while analyzing 11 patients without known consanguinity, detecting genome-wide AOH levels of between 3.5 and 8.4 percent.
"We were concerned that these AOH regions could be false-positives, so we analyzed the same patients on a high-density SNP platform," Rudd said. For the comparison, EGL selected Affymetrix's 2.6-million-feature CytoScan HD, which contains 750,000 SNPs and 1.9 million oligo probes.
Testing of the same 11 samples with CytoScan yielded AOH levels of between 0 and 0.66 percent.
After Rudd and colleagues discussed the findings at an American College of Medical Genetics meeting, they received "a lot of feedback from other clinical laboratories using SNP arrays" and decided to partner with Boston Children's Hospital to analyze CNV and AOH using a third, custom, mid-density Agilent array containing 180,000 oligos and 30,000 SNPs.
Altogether, they evaluated 50 samples using the three platforms. According to Rudd, the Agilent array also detected very low levels of AOH in these 11 samples, leading her lab to conclude that the original AOH calls on the OGT platform were false-positives, and prompting the paper's authors to advocate the use of higher-density SNP arrays for detection of AOH.
As they wrote in the paper, while lower-density arrays such as OGT's CytoSure ISCA UPD are "suitable" for UPD detection, copy-number plus single-nucleotide polymorphism arrays with 30,000 or fewer unique SNP probes "miscall" AOH regions due to identity by descent.
Rudd said that EGL ultimately decided to adopt CytoScan for copy number and SNP analysis, but that it still uses a 60,000-oligonucleotide OGT array for copy number only testing. "After doing the array comparison, it was clear that CytoScan had the best SNP density for AOH detection," she said.
In the paper, the authors noted that CytoScan detected some small regions of AOH that the custom Agilent array missed. Still, they said that the absence of these regions from the Agilent-based analysis "did not significantly alter the percentage autosomal AOH" and did not change the interpretation of the patient result. Therefore, they reported in the paper that both the Affy and Agilent arrays "reliably detect AOH due to identity by descent."
A Need for Comparison
Emory and Boston Children's Hospital's evaluation of the three platforms occurred in the past few years, a time when Agilent, as well as companies that sell Agilent-manufactured chips, such as OGT, began adding SNP content to their CGH arrays (BAN 4/17/2012).
During the same time period Affy introduced CytoScan, which contained more oligo content for detecting copy number changes and replaced the SNP 6.0 Array as its main offering for the cytogenetics community (BAN 5/10/2011).
One of the reasons that Agilent and its clients began adding SNP content, and Affy improved its CNV detection capabilities, was to provide labs with products that could detect both CNVs and homozygosity, as many labs had been using two tools to interrogate the same samples: Agilent-made CGH arrays for CNV detection, and Affy and Illumina SNP arrays to detect stretches of homozygosity.
Differences in array manufacturing capabilities allowed Affy to introduce an array that had more than twice the number of features of Agilent's highest density array product, and also meant that Agilent and its client firms had to be more selective in the SNP content they included on their arrays.
"Given the differences in SNP density across the three platforms, there was a real need to evaluate their performance in AOH detection," said Rudd. She said that other clinical labs told her that they too were "struggling" to decide on the "best array" for CNV and AOH detection.
"Considering the cost of equipment and analysis software, choosing an array platform is a major investment for any clinical laboratory," Rudd said of the rationale for the study. "We wanted to share our experience with other groups so they could benefit from the lessons we learned the hard way," she said.
Rudd cautioned, though, that while arrays with higher SNP density may be "the best" for detecting AOH, that does not mean that all labs should adopt CytoScan or similar arrays.
"All three array platforms performed well in detection of CNVs, UPD, and chromosome mosaicism," said Rudd. "Chromosome microarrays were originally designed to pick up copy number changes, not long stretches of homozygosity," she said. "So if a laboratory is only looking for copy number analysis, the three platforms are comparable."
In the paper, the authors also argued that it is possible that analysis paramaters on lower-density arrays could be altered to improve AOH detection. By increasing the stringency of the loss-of-heterozygosity score in OGT's software, the authors claimed that they were able to reduce the number of false positives and lower the percentage of autosomal AOH to levels expected in a normal population. At the same time, raising the LOH cutoff inflated the number of missed, or false-negative, AOH regions.
Because of this, they wrote that there is a "significant trade-off" between reducing false positives in regions that are not AOH and increasing false negatives in true stretches of AOH" when choosing lower-density array software settings.