In fact, SNP chips are “extremely unreliable for genotyping very rare pathogenic variants,” and a positive result for such a variant “is more likely to be wrong than right,” researchers reported in the BMJ.
The authors explained that SNP chips are “DNA microarrays that test genetic variation at many hundreds of thousands of specific locations across the genome.” Although SNP chips have proven accurate in identifying common variants, past reports have suggested that SNP chips perform poorly for genotyping rare variants.
To gain more insight, Caroline Wright, PhD, of the University of Exeter (England) and colleagues conducted a large study.
The researchers analyzed data on 49,908 people from the UK Biobank who had SNP chip and next-generation sequencing results, as well as an additional 21 people who purchased consumer genetic tests and shared their data online via the Personal Genome Project.
The researchers compared the SNP chip and sequencing results. They also selected rare pathogenic variants in BRCA1 and BRCA2 for detailed analysis of clinically actionable variants in the UK Biobank, and they assessed BRCA-related cancers in participants using cancer registry data.
Largest evaluation of SNP chips
SNP chips performed well for common variants, the researchers found. Sensitivity, specificity, positive-predictive value, and negative-predictive value all exceeded 99% for 108,574 common variants.
For rare variants, SNP chips performed poorly, with a positive-predictive value of 16% for variants with a frequency below 0.001% in the UK Biobank.
“The study provides the largest evaluation of the performance of SNP chips for genotyping genetic variants at different frequencies in the population, particularly focusing on very rare variants,” Dr. Wright said. “The biggest surprise was how poorly the SNP chips we evaluated performed for rare variants.”
Dr. Wright noted that there is an inherent problem built into using SNP chip technology to genotype very rare variants.
“The SNP chip technology relies on clustering data from multiple individuals in order to determine what genotype each individual has at a specific position in their genome,” Dr. Wright explained. “Although this method works very well for common variants, the rarer the variant, the harder it is to distinguish from experimental noise.”
False positives and cancer: ‘Don’t trust the results’
The researchers found that, for rare BRCA variants (frequency below 0.01%), SNP chips had a sensitivity of 34.6%, specificity of 98.3%, negative-predictive value of 99.9%, and positive-predictive value of 4.2%.
Rates of BRCA-related cancers in patients with positive SNP chip results were similar to rates in age-matched control subjects because “the vast majority of variants were false positives,” the researchers noted.
“If these variants are incorrectly genotyped – that is, false positives detected – a woman could be offered screening or even prophylactic surgery inappropriately when she is more likely to be at population background risk [for BRCA-related cancers],” Dr. Wright said.
“For very-rare-disease–causing genetic variants, don’t trust the results from SNP chips; for example, those from direct-to-consumer genetic tests. Never use them to guide clinical action without diagnostic validation,” she added.
Heather Hampel, a genetic counselor and researcher at the Ohio State University Comprehensive Cancer Center in Columbus, agreed.
“Positive results on SNP-based tests need to be confirmed by medical-grade genetic testing using a sequencing technology,” she said. “Negative results on an SNP- based test cannot be considered to rule out mutations in BRCA1/2 or other cancer-susceptibility genes, so individuals with strong personal and family histories of cancer should be seen by a genetic counselor to consider medical-grade genetic testing using a sequencing technology.”
Practicing oncologists can trust patients’ prior germline genetic test results if the testing was performed in a cancer genetics clinic, which uses sequencing-based technologies, Ms. Hampel noted.
“If the test was performed before 2013, there are likely new genes that have been discovered for which their patient was not tested, and repeat testing may be warranted,” Ms. Hampel said. “A referral to a cancer genetic counselor would be appropriate.”
Ms. Hampel disclosed relationships with Genome Medical, GI OnDemand, Invitae Genetics, and Promega. Dr. Wright and her coauthors disclosed no conflicts of interest. The group’s research was conducted using the UK Biobank and the University of Exeter High-Performance Computing, with funding from the Wellcome Trust and the National Institute for Health Research.