Misannotation of multiple-nucleotide variants risks misdiagnosis

This is the published version [version 2; peer review: 2 approved]. Available on open access from F1000Research via the DOI in this recordData availability: Underlying data: Simulated MNV data is available at https://github.com/rdemolgen/MNV-test-data Archived simulated MNV data at time of publication: http://doi.org/10.5281/zenodo.3375579 License: GNU General Public License v3.0 The dataset of 1447 samples previously sequenced cannot be shared due to patient confidentiality issues, as the genotype data could be used to identify individuals and so cannot be made openly available. Requests for access to the anonymised data by researchers will be considered following an application to the Genetic Beta Cell Research Bank (https://www.diabetesgenes.org/current-research/genetic-beta-cell-research-bank/) with proposals reviewed by the Genetic Data Access Committee.Multiple Nucleotide Variants (MNVs) are miscalled by the most widely utilised next generation sequencing analysis (NGS) pipelines, presenting the potential for missing diagnoses. These variants, which should be treated as a single insertion-deletion mutation event, are commonly called as separate single nucleotide variants. This can result in misannotation, incorrect amino acid predictions and potentially false positive and false negative diagnostic results. Using simulated data and re-analysis of sequencing data from a diagnostic targeted gene panel, we demonstrate that the widely adopted pipeline, GATK best practices, results in miscalling of MNVs and that alternative tools can call these variants correctly. The adoption of calling methods that annotate MNVs correctly would present a solution for individual laboratories, however GATK best practices are the basis for important public resources such as the gnomAD database. We suggest integrating a solution into these guidelines would be the optimal approach.Wellcome TrustNewlife Foundation for Disabled Childre