Chip-based DNA testing wrong more than right for very rare variants

By Matthew Sparkes

The type of chips used in some consumer DNA tests struggles to separate rare mutations from experimental noise, and a positive result for a rare cancer-causing gene variant is more likely to be wrong than right. Results for common variants are better, though.

Single-nucleotide polymorphism (SNP) chips are a cheaper way to map genomes than full genetic sequencing. Using grids of hundreds of thousands of beads that react to specific gene variants by glowing in different colours, these chips can map whichever parts of the whole genome their designers choose. They are accurate for common gene variants of the type that can trace ancestry or that are tied to the risk of type 2 diabetes, but with rarer gene variants, it becomes far harder to separate data from noise.

Caroline Wright at the University of Exeter, UK, and her colleagues have now shown that SNP chips can pinpoint common variants of two genes linked to cancer called BRCA1 and BRCA2 with 99 per cent accuracy. But with rarer variants it was more likely to show a false positive or false negative than be accurate.

Advertisement

The teams’s study used data from the UK Biobank, which began collecting DNA samples for research from 500,000 people in 2006. All of those samples are processed with SNP chips, but Wright’s team looked at a subset of almost 50,000 people (55 per cent of whom were female) who had also been tested with full genetic sequencing.

Variants in the BRCA genes can be passed down from either parent and dozens are linked to cancer. People incorrectly told they have cancer-causing variants face additional screening and potentially unnecessary surgery. Those incorrectly told that they don’t have the gene variants may miss out on preventative interventions.

The team found that only 17 per cent of the 189 people who had received a positive BRCA result from SNP tests were shown to have those variants by genome sequencing. Of the 45,678 people given an all-clear by SNP, 65 were shown to actually be positive. A further 70 people who had negative results from the SNP data were revealed by sequencing to have one or more of 43 cancer-causing variants.

The team also investigated rare variants among people who had submitted their own DNA testing results from various sources to the Personal Genome Project UK for open research. Only 21 subjects had submitted both SNP and sequencing data, but all of them had been incorrectly labelled by the SNP chips in regard to BRCA.

Wright says that the data reveals SNP chips to be “extremely poor” for detecting rare gene variants. In fact, for an individual person, a positive result for a very rare pathogenic variant is more likely to be wrong than right.

She warns that anyone receiving a negative result for cancer-causing variants from a consumer DNA test shouldn’t be reassured that their risk of hereditary cancer is low. Nor should they seek medical interventions after a positive result without having the data validated by sequencing by their doctor. “It just is not the right technology for the job,” she says.

Most consumer DNA testing relies on SNP chips, although the ones used now have benefitted from several years of additional research and development compared with those looked at in the UK Biobank data by this study. Many firms advertise that their tests can screen for BRCA variants but often look for only a small number of the more prevalent mutations.

Journal reference: BMJ, DOI: 10.1136/bmj.n214