Artificial Intelligence Approach for Variant Reporting.

Purpose: Next-generation sequencing technologies are actively applied in clinical oncology. Bioinformatics pipeline analysis is an integral part of this process; however, humans cannot yet realize the full potential of the highly complex pipeline output. As a result, the decision to include a variant in the final report during routine clinical sign-out remains challenging. Methods: We used an artificial intelligence approach to capture the collective clinical sign-out experience of six board-certified molecular pathologists to build and validate a decision support tool for variant reporting. We extracted all reviewed and reported variants from our clinical database and tested several machine learning models. We used 10-fold cross-validation for our variant call prediction model, which derives a contiguous prediction score from 0 to 1 (no to yes) for clinical reporting. Results: For each of the 19,594 initial training variants, our pipeline generates approximately 500 features, which results in a matrix of > 9 million data points. From a comparison of naive Bayes, decision trees, random forests, and logistic regression models, we selected models that allow human interpretability of the prediction score. The logistic regression model demonstrated 1% false negativity and 2% false positivity. The final models' Youden indices were 0.87 and 0.77 for screening and confirmatory cutoffs, respectively. Retraining on a new assay and performance assessment in 16,123 independent variants validated our approach (Youden index, 0.93). We also derived individual pathologist-centric models (virtual consensus conference function), and a visual drill-down functionality allows assessment of how underlying features contributed to a particular score or decision branch for clinical implementation. Conclusion: Our decision support tool for variant reporting is a practically relevant artificial intelligence approach to harness the next-generation sequencing bioinformatics pipeline output when the complexity of data interpretation exceeds human capabilities.

Using large sequencing data sets to refine intragenic disease regions and prioritize clinical variant interpretation.

PURPOSE: Classification of novel variants is a major challenge facing the widespread adoption of comprehensive clinical genomic sequencing and the field of personalized medicine in general. This is largely because most novel variants do not have functional, genetic, or population data to support their clinical classification. METHODS: To improve variant interpretation, we leveraged the Exome Aggregation Consortium (ExAC) data set (N = ~60,000) as well as 7,000 clinically curated variants in 132 genes identified in more than 11,000 probands clinically tested for cardiomyopathies, rasopathies, hearing loss, or connective tissue disorders to perform a systematic evaluation of domain level disease associations. RESULTS: We statistically identify regions that are most sensitive to functional variation in the general population and also most commonly impacted in symptomatic individuals. Our data show that a significant number of exons and domains in genes strongly associated with disease can be defined as disease-sensitive or disease-tolerant, leading to potential reclassification of at least 26% (450 out of 1,742) of variants of uncertain clinical significance in the 132 genes. CONCLUSION: This approach leverages domain functional annotation and associated disease in each gene to prioritize candidate disease variants, increasing the sensitivity and specificity of novel variant assessment within these genes.Genet Med advance online publication 22 September 2016.

Improving hearing loss gene testing: a systematic review of gene evidence toward more efficient next-generation sequencing-based diagnostic testing and interpretation.

PURPOSE: With next generation sequencing technology improvement and cost reductions, it has become technically feasible to sequence a large number of genes in one diagnostic test. This is especially relevant for diseases with large genetic and/or phenotypic heterogeneity, such as hearing loss. However, variant interpretation remains the major bottleneck. This is further exacerbated by the lack in the clinical genetics community of consensus criteria for defining the evidence necessary to include genes on targeted disease panels or in genomic reports, and the consequent risk of reporting variants in genes with no relevance to disease. METHODS: We describe a systematic evidence-based approach for assessing gene-disease associations and for curating relevant genes for different disease aspects, including mode of inheritance, phenotypic severity, and mutation spectrum. RESULTS: By applying this approach to clinically available hearing loss gene panels with a total of 163 genes, we show that a significant number (45%) of genes lack sufficient evidence of association with disease and thus are expected to increase uncertainty and patient anxiety, in addition to intensifying the interpretation burden. Information about all curated genes is summarized. Our retrospective analysis of 539 hearing loss cases tested by our previous OtoGenomeV2 panel demonstrates the impact of including genes with weak disease association in laboratory wet-bench and interpretation processes. CONCLUSION: Our study is, to our knowledge, the first to highlight the urgent need for defining the clinical validity of gene-disease relationships for more efficient and accurate clinical testing and reporting.Genet Med 18 6, 545-553.