Making Sense of the Genome Remains a Work in Progress.

DNA sequence information has become increasingly easier and less expensive to obtain, but interpretation remains a challenge. The vast amount of variation present in the human genome reveals the scope of the problem. For instance, analysis of all protein-coding regions of the genome (an exome analysis) in 50 726 individuals found a median of more than 20 000 gene variants per person, most of them rare and hundreds not previously identified.1 Because most genomic variation has little or no observable effect on human health, laboratories must undertake careful evaluation to determine the pathogenicity of variants identified by genomic testing. Standards promulgated in 2015 call for the consideration of several different types of evidence, including clinical observation, family segregation, population prevalence, and computational and functional data.2 However, the evidence for most variants is limited, so results regarding pathogenicity are often uncertain. One common outcome is a gene variant of unknown clinical significance. Other variants are classified as likely pathogenic or benign (defined as ≥90% likelihood), yielding a 5-point classification scheme: pathogenic; likely pathogenic; variant of unknown significance; likely benign; and benign.2 Over time, a gene variant may be reclassified, changing, for example, from variants of unknown significance to likely pathogenic or vice versa. But how often does this happen and under what time frame? The report by Mersch et al3 in this issue of JAMA offers some answers to these critical questions. The study by Mersch et al reports findings from a 10-year experience with hereditary cancer genetic testing at a major national laboratory. Results from 2 types of testing were reported. In 1 156 522 patients, a single gene or small number of genes linked to a particular genetic condition was assessed. In 304 664 patients, a multigene panel was used, testing for numerous different genes associated with different inherited cancer syndromes. About half of the patients were of European descent and 60% had been affected with cancer. Most were women, presumably reflecting the predominance of genomic testing related to breast and ovarian cancer. The laboratory used an automated system to identify new evidence relevant to variant classification on an ongoing basis. On initial testing, 5.4% of patients had at least 1 variant classified as pathogenic or likely pathogenic, and 5.8% had at least 1 variant of unknown significance in the absence of an accompanying pathogenic or likely pathogenic variant. In initial classification of the 44 777 unique variants identified from testing, 9112 were designated as pathogenic or likely pathogenic, 8995 as benign or likely benign, and 26 670 as variants of unknown significance. Over the study period, 0.2% of benign or likely benign variants were upgraded (to a variants of unknown significance or to a pathogenic category) and 0.7% of pathogenic or likely pathogenic variants were downgraded (to a variants of unknown significance or to a benign category). Among variants of unknown significance, 7.7% of unique variants were reclassified (affecting 24.9% of clinical reports, because many of these variants were observed in ≥1 person). For reclassified variants of unknown significance, 91.2% (1867 of 2048) were downgraded to a benign classification, with the remainder upgraded to a pathogenic category. Even though this study is limited to the experience of a single laboratory and to tests focused on cancer, it provides valuable data for clinicians and patients. Most important, it sheds light on the problems with variants of unknown significance. Among those reclassified, the vast majority proved to be benign, a reassuring result. However, less than 10% of unique variants of unknown significance were reclassified during the 10year course of the study. So for most people with variants of unknown significance, there is little hope of timely resolution of the uncertainty generated by the result. The prospects are lower when the variant in question is rare, or when resources to evaluate the variant are limited. For example, detailed family studies may clarify whether a rare variant of unknown significance is pathogenic or benign, but the procedures for such studies, which involve identifying and testing several family members, are typically beyond the scope of current practice.4 Given the sheer number of gene variants that will be identified with increasing use of genomic testing, the problem posed by variants of unknown significance is substantial. Downgrading of pathogenic variants and upgrading of benign variants were fortunately rare in the study by Mersch et al. However, misclassification may be higher in practice. Recent studies have documented that different laboratories may produce discordant interpretations of the same variant.5,6 One study of 9 laboratories found that the concordance rate for interpretation of 99 variants spanning the full range of classification categories was only 34%.6 After consensus discussions, concordance increased to 71%,6 underscoring the need for judgment, and, therefore, the potential for disagreement in interpreting data on variant pathogenicity.7 In addition, Mersch et al focused on cancer genetics, the clinical domain with the most extensive clinical experience. Genetic testing in clinical areas other than cancer may have a greater likelihood of generating both variants of unknown significance and erroneous determinations of pathogenicity. The challenges of variant interpretation are well recognized in the field, and concerted efforts are under way to improve consistency and accuracy.2,7,8 These include the Related article page 1266 Opinion