Ximmer: A System for Improving Accuracy and Consistency of CNV Calling from Exome Data

Abstract

Detection of copy number variation (CNVs) is a challenging but highly valuable application of exome and targeted high throughput sequencing (HTS) data. While there are dozens of CNV detection methods available, using these methods remains challenging due to variable accuracy both across different data sets and within the same data set with different methods. We propose that extracting good results from CNV detection on HTS data requires a systematic approach involving rigorous quality control, adjustment of method parameters and calibration of confidence measures for filtering results. We present Ximmer, a tool which supports an end to end process for applying these procedures including a simulation framework, CNV detection analysis pipeline, and a visualisation and curation tool which enables interactive exploration of CNV results. We apply Ximmer to perform a comprehensive evaluation of CNV detection on four data sets using four different detection methods, representing one of the most comprehensive evaluations to date. Ximmer is open source and freely available at http://ximmer.org (example results are viewable at http://example.ximmer.org).