An Augmented Exome Providing Accurate Structural Variant Detection. A. Patwardhan, S. Chervitz, M. Li, J. Harris, G. Bartha, D. Newburger, M. Pratt, S. Garcia, J. Tirch, N. Leng, C. Haudenschild, S. Luo, D. Church, J. West, R. Chen Personalis, Inc., Menlo Park, Ca.

   Whole-Exome Sequencing (WES) has proven to be an efficient tool in identifying common and rare disease-associated variants in the protein-coding region of the genome. This has had a direct impact on the clinical setting, where WES has provided definitive diagnoses to patients affected by Mendelian Disorders, even in cases where other diagnostic tests have failed. While WES theoretically provides coverage of all potential disease-associated variants within the exome, the majority of published exome studies report only the detection of small variants (SNVs, InDels). The false-negative rate for Structural Variations (SVs), a class of large (>50bp) variants that include copy number variations, is thought to remain high given their established or expected importance in disease phenotype. The size and complexity of SVs increases the likelihood that they will encompass a region of variation not well-captured by standard WES, partially due to the targeted nature of the capture methods, incomplete gene coverage, and the limited size of coding sequence.
   We describe a novel approach that combines an augmented exome assay with novel informatics to address many of these technical challenges and improve SV calling accuracy. The augmented exome improves coverage over all biomedically interpretable genes compared to standard WES assays and extends coverage to detect large SVs genome wide. Using read-depth information, SV detection is performed concurrently in exome-targeted regions and genome-wide, with corrections for non-uniform coverage when appropriate.
   SV detection rates were compared among a set of over 40 samples known to harbor pathogenic deletions and duplications using augmented exome and standard WES approaches. Additionally, the accuracy of our approach relative to standard WES is estimated by comparing SVs detected in a reference sample to a gold set of SVs developed in-house. This gold-set is derived from a deeply sequenced 16-member pedigree where SVs in the reference sample are vetted by breakpoint, inheritance state, and variant type consistency among related pedigree members. Using both the known samples and the gold-set, we demonstrate increased accuracy in detecting SVs with the augmented exome over a range of SV sizes (1KB- >1MB).

You may contact the first author (during and after the meeting) at