President's Corner - Quantitative Interpretation of Sequence Variants, Part 2
Published on
In my previous President’s corner, I outlined the rationale for quantitative interpretation of sequence variants and presented our quantitative definitions of the five interpretation categories (Pathogenic, Likely Pathogenic, Uncertain, Likely Benign and Benign). In this Issue, I discuss in more detail two quantitative approaches with results and with limitations and pitfalls.
Our first approach is simply the comparison of the frequency of a variant in affected individuals (cases) versus controls. This is certainly not a novel approach, but is still quite powerful and useful. An example is shown in Figure 1 for dominantRYR1 variants responsible for Malignant Hyperthermia. For cases, we used only clearly affected probands tested at PreventionGenetics. For controls, we used ExAC allele frequencies combined for all populations. p values were calculated using Fisher’s exact test. Clear differences can be seen in the p values for between Pathogenic and Benign variants (interpreted using ACMG guidelines in the absence of quantitative assistance). We have also shown that this approach works well for recessive disease.
All interpretation approaches have limitations and pitfalls. One significant limitation of comparison of case and control allele frequencies is that the variant of interest must be observed multiple times. A single observation is very likely insufficient. We used a minimum of two cases in our analyses, but three or more cases should provide better discrimination and more accurate p values. Potential pitfalls of this approach include that the populations used for cases and controls need to be the same, and that errors and omissions in the control allele frequency databases (which do exist) must be avoided.
Our second quantitative approach applies only to autosomal recessive disorders. Data for the fraction of times the variant of interest was accompanied by a second pathogenic or likely pathogenic variant for Autosomal Recessive Polycystic Kidney Disease (PKHD1 gene) are shown in the Figure 2. With this approach, there is also generally clear distinction between pathogenic and benign variants. As before, analysis was restricted to variants observed in at least two affected individuals. Some of the likely pathogenic and uncertain variants which were seen every time with second pathogenic variants are candidates for pathogenic designation. Variants with intermediate fractions may have incomplete penetrance.
Note that for most recessive disorders, the true fraction seen with second pathogenic variants will not be 100% because copy number variants and other variants not detectable by sequencing will be present. Also, less commonly, a proband may be only a heterozygous carrier for the variant of interest. Ideally therefore, known pathogenic variants should be used as controls for each disorder analyzed. For both of the quantitative approaches described here, data can come from either testing labs or from the literature. We strongly recommend that such quantitative approaches be used as a first line approach for interpretation wherever possible. We also recommend that all non-quantitative evidence also be carefully examined for each variant. I hope that the clinical genetics community will see a rapid move toward quantitative interpretation approaches and away from the highly subjective approaches outlined in the current ACMG Guidelines. Our work is an important step in this process, but just a first step. More powerful statistical methods very likely can be employed and additional quantitative approaches are also likely.
-James L. Weber, PhD
Founder & President, PreventionGenetics