A precision metric for clinical genome sequencing

Abstract

A requisite precondition for the application of next-generation sequencing to clinical medicine is the ability to confidently call genotype at each coding/splicing position of every gene of interest. Current gold standard technologies, such as Sanger sequencing and microarrays, allow confident identification of the genomic origin of the DNA of interest. A commonly used minimum standard for the adoption of new technology in medicine is non-inferiority. We developed a metric to quantify the extent to which current sequencing technologies reach this clinical grade reporting standard. This metric, the rationale for which we present here, is defined as the absolute number of base pairs per gene not callable with confidence, as specified by the presence of 20 high quality (Q20) bases from uniquely mapped (mapq>0) reads per locus. To illustrate the utility of this metric, we apply it across data from several commercially available clinical sequencing products. We present specific examples of coverage for genes known to be important for clinical medicine. We derive data from a variety of platforms including whole genome sequencing (Illumina Hiseq and X chemistry) and exome capture (including medically optimized capture from Agilent, Baylor Clinical Lab, and Personalis). We observe that compared to whole genomes (with {small tilde}30x average coverage), augmented exomes perform far better for known disease causing genes, but less well for other genes and in untranslated regions. Increasing whole genome coverage improves this discrepancy with an average coverage of {small tilde}45x representing the cross over point where performance equals that of exome capture for disease causing genes. A combination of some genome-wide coverage and augmented exon coverage may offer the most cost effective solution for clinical grade genome sequencing today. In summary, this coverage metric provides transparency regarding the current state of next-generation sequencing for clinical medicine and will inform genotype interpretation, technology improvement, and sequencing platform choices for physicians and laboratories. We provide an application on precision.fda.gov (Coverage of Key Genes app) to calculate this metric.