Evaluating in silico approaches to improving missense variant interpretation in genomic medicine

Guidelines published by the ACMG/AMP in 2015 and ACGS in 2020 provide a framework for the assessment and classification of novel variants identified through genetic testing; providing consistency and transparency to the variant classification process. Using variant datasets derived either from online databases, such as ClinVar or gnomAD, or novel variants identified through research and diagnostic panel and exome sequencing, a number of different aspects of the ACMG/ACGS guidelines are assessed for performance. Five key aspects are covered: (1) the use of is silico missense predictors (2) the use of co-localising alternative pathogenic variants (3) variants present in pseudo-population databases, and (4) the use of genetic constraint data (5) the use of multiple complementary in silico tools to assess the likely functional impact of a novel 67Kb duplication identified in 5 neonates; a variant type that is typically difficult to characterise and classify through standardised variant classification guidelines. A novel concept is introduced: the use of meta-positions, whereby the availability of data is increased through the use of information from regions considered to be functionally-equivalent to that being assessed. Internal inconsistencies in evidence weighting are identified and recommendations made over the usage and weighting of specific tools and algorithms; the application of evidence at different weighting based on algorithm score, and the use of LR+ to determine weighting within the framework. Together, this work seeks to improve the application of the variant classification guidelines through evidence-based weighting of criteria, and simplify the process by eliminating the burden of performing recursive variants assessments. The use of additional evidence to support the benignity of a variant, not currently implemented in the guidelines, can be used to exclude variants from the need for manual assessment.