Utility of splicing prediction tools in variant interpretation
Multiple in silico tools have been developed to predict the impact of spliceogenic variants (Table 1), and such prediction is an important component of variant curation and interpretation processes for Mendelian disease genes. Several studies have assessed the utility of prediction tools in interpretation of variants in hereditary breast and ovarian cancer genes. These include: clinical calibration of the MaxEntScan (MES) tool to estimate the prior probability of pathogenicity of genetic variation in BRCA1 and BRCA2 due to impact on native donor and acceptor motifs, or the creation of exonic de novo donor sites (Vallée et al., 2016); assessment of the sensitivity and specificity of different MES thresholds to predict aberrant splicing using experimentally validated spliceogenic variants in BRCA1 ,BRCA2 , MLH1, MSH2, MSH6 and PMS2 (Shamsani et al., 2018); and the combined use of MES and Splice Site Finder-like, trained and validated using in vitro mRNA data to improve in silico prediction of spliceogenic variants in donor and acceptor splice site motifs (Leman et al., 2018). The combined MES and Splice Site Finder-like analysis pipeline has been previously proposed as a prioritization method for splicing analysis of BRCA1 andBRCA2 variants of uncertain significance (VUSs) (Houdayer et al., 2012). These studies have shown the reliability of bioinformatic tools in predicting spliceogenic variants in the donor and acceptor splice site motifs, especially those that disrupt the highly conserved dinucleotides at the 3’ (AG) and 5’ splice sites (GT).
In contrast, predictors of variant effects on exonic SREs or BP sites currently perform poorly (see below), which limits their utility to inform variant classification in routine diagnostics. There is currently no prediction tool specifically designed for pseudoexon-activating variants. In the following sections, we discuss: i) the spliceogenic variants outside of the donor and acceptor splice site motifs; ii) the current tools used to predict their effects on splicing and their predictive performance; and iii) combined strategies using functional studies and in silico tools to prioritize variants for confirmatory splicing assays.