Research
Use of SNP chips to detect rare pathogenic variants: retrospective, population based diagnostic evaluation
BMJ 2021; 372 doi: https://doi.org/10.1136/bmj.n214 (Published 16 February 2021) Cite this as: BMJ 2021;372:n214Linked Opinion
SNP chips perform poorly for detecting very rare genetic variants
Re: Use of SNP chips to detect rare pathogenic variants: retrospective, population based diagnostic evaluation
Dear Editor,
As clinical and societal demands of medical genetic testing rise, eyes turn towards genotyping arrays as a high-throughput and cost-effective alternative for sequencing methods. In contrast to common variants, detection of the rare variants, increasingly added to arrays, is less explored, which is why we much welcome studies like from Weedon et al. [1].
However, Weedon et al. retrospectively compared rare variants detected by array and by sequencing for ~50,000 subjects from UK Biobank (UKBB) and for 21 customers of 23andMe and conclude that arrays are “extremely unreliable for genotyping very rare (pathogenic) variants”. We were surprised by several aspects of this BMJ publication, and argue that the main conclusion is wrong and arrays are in fact very well suited to detect rare variants, including supposedly pathogenic variants, if the right QC and validation procedures are applied.
Firstly, the UKBB array stems from ~2012 and was not intended for rare variant detection, or for diagnostic reporting of such variants. The 23andMe arrays are more recent, but details on data collection and processing are lacking, and this concerns a small sample set. Therefore, these are suboptimal datasets to draw conclusions on rare variant genotyping and reporting.
Secondly, it is customary to validate and optimize the array genotyping procedure, especially when particular classes of variants are involved, such as e.g. pharmacogenetic markers, or markers for HLA or ancestry, and of course rare variants. In particular rare-variant specific clustering can be optimized through algorithms like zCall [2] or manual evaluation of cluster plots by exclusion of probes with confounding factors (e.g., a-specific binding of adjacent variants such as benign BRCA1/2 polymorphisms triggering probes intended for pathogenic variants) and by validation studies confirming that the assay operates as intended. We were therefore disappointed that none of these well-known options were applied in the study by Weedon et al. [1]. We note that also for genotyping protocols based on next generation sequencing data (which the authors take as “golden standard”), such optimization has to be done as the authors acknowledge by noting that “…sites that were not well sequenced…” needed to be excluded.
Thirdly, we point towards publications which have successfully used arrays to detect rare variants, such as through the Illumina’s HumanExome Beadchip [3,4], which was actually developed for this purpose, or more recent customized arrays such as used by the million veterans project [5] or by our own group [6]. In fact, using a similar optimized array procedure in a blinded study of 240 breast cancer patients with clinically confirmed pathogenic variants (mostly in BRCA1 and BRCA2), we could correctly identify the causal variant in >95% of carriers and only observed false positive signals in <1% of samples (manuscript in preparation).
Thus, in contrast to Weedon et al., we conclude that arrays are suitable for genotyping (very) rare genetic variants, both pathogenic and non-pathogenic. Given their much lower pricing, design flexibility and highly standardized use, arrays offer, when data is analyzed appropriately, a very attractive genetic analysis tool for many different purposes, including clinical ones.
Best regards,
Jeroen van Rooij1, Annemieke Verkerk1,2, Jard de Vries2, Linda Broer2, Joyce van Meurs1,2, André Uitterlinden1,2
1 Genetic Laboratory, Department of Internal Medicine, Erasmus MC, Rotterdam, the Netherlands
2 Human Genomics Facility, Erasmus MC, Rotterdam, the Netherlands
References:
1. Weedon et al. Use of SNP chips to detect rare pathogenic variants: retrospective, population-based diagnostic evaluation. BMJ 2021;372:n214; doi:10.1136/bmj.n214.
2. Goldstein et al. zCall: a rare variant caller for array-based genotyping: genetics and population analysis. Bioinformatics 2012, Oct 1;28(19):2543-5. doi: 10.1093/bioinformatics/bts479. Epub 2012 Jul 27.
3. Dekker et al. Exome array analysis of rare and low frequency variants in amyotrophic lateral sclerosis. Scientific Reports, 2019.
4. Grove et al. Best practices and joint calling of the HumanExome BeadChip: the CHARGE Consortium. PLoS One. 2013 doi: 10.1371.
5. Hunter-Zinck et al. Genotyping array design and data quality control in the million veteran program. AJHG 106, 535-548, 2020.
6. Suratannon et al. Rapid low-cost microarray-based genotyping for genetic screening in primary immunodeficiency. Front Immunol; Apr 2020; doi:10.3389/fimmu.2020.00614.
Competing interests: No competing interests