34. Identifying challenges in variant normalization

Genomic medicine relies on collecting information from multiple sources to make optimal therapeutic and diagnostic decisions for the patient. However, integration of this information, at the time of variant interpretation, is a major bottleneck. Challenges including inconsistent formats (e.g. HGVS and SPDI), and variant contexts (genome and proteome), increase the time and effort required to formulate and communicate a complete variant interpretation. To more clearly understand inter-resource differences in variant representation, variants from the Clinical Interpretations of Variants in Cancer (CIViC), Molecular Oncology Almanac (MOAlmanac), and ClinVar were evaluated using a normalization protocol. Of the variants in the knowledgebases, ∼53% of the CIViC, ∼42% of the MOAlmanac, and ∼99% of ClinVar variants were successfully normalized. We categorically assessed remaining variants for which normalization methods are still needed, and analyzed these for clinical impact. Gene fusion (e.g. 'ALK G1202R') and region-defined variant categories (e.g. '3′ UTR Mutations') respectively constitute 16% and 10% of all the variants that were not normalized, and were found to hold the greatest potential clinical impact. Additionally, fusion variants are responsible for ∼25% of all of the evidence items associated with not normalized variants from CIViC and MOAlmanac, illustrating the weight that fusion variants carry in the overall group. The Variation Normalizer is an open-source toolkit and is available for use with independent data sets to facilitate precise matching of evidence from knowledgebases to genomic variant data.