Comparison of literature mining tools for variant classification: Through the lens of 50 RYR1 variants

Abstract

Purpose

The American College of Medical Genetics and Genomics and the Association for Molecular Pathology have outlined a schema that allows for systematic classification of variant pathogenicity. Although gnomAD is generally accepted as a reliable source of population frequency data and ClinGen has provided guidance on the utility of specific bioinformatic predictors, there is no consensus source for identifying publications relevant to a variant. Multiple tools are available to aid in the identification of relevant variant literature, including manually curated databases and literature search engines. We set out to determine the utility of 4 literature mining tools used for ascertainment to inform the discussion of the use of these tools.

Methods

Four literature mining tools including the Human Gene Mutation Database, Mastermind, ClinVar, and LitVar 2.0 were used to identify relevant variant literature for 50 RYR1 variants. Sensitivity and precision were determined for each tool.

Results

Sensitivity among the 4 tools ranged from 0.332 to 0.687. Precision ranged from 0.389 to 0.906. No single tool retrieved all relevant publications.

Conclusion

At the current time, the use of multiple tools is necessary to completely identify the literature relevant to curate a variant.