Conclusions
In this paper, we have presented functional evidence for spliceogenic
variants that are generally overlooked in clinical genetic testing
and/or reporting, including: variants that affect SREs, abrogate BP
sites, or activate pseudoexons. Bioinformatic analysis considering
variant effects at the mRNA level may help prioritize likely functional
variants currently annotated as (likely) benign or VUS for additional
functional and clinical analyses. Further, clinical diagnostic
laboratories may need to consider expanding their sequencing coverage
and/or variant annotation to include BP window and deep intronic regions
to detect additional pathogenic intronic variants, particularly when
strongly indicated by patient presentation. However, improving the low
performance of current predictors is a challenge due to the limited size
of experimentally validated training data. Clearly, experimental studies
that assess variants outside of the donor and acceptor splice site
motifs for splicing mechanisms are needed to further calibrate
algorithms, and to improve prediction of variant effect. We have shown
that results from a published large-scale saturation genome editing
experiment can be used to map SREs, to assess the performance of
bioinformatic predictors, and inform development of a prioritization
workflow to detect variants that impact SREs. As more such data become
available, we anticipate that the expansion of training datasets will
lead to improvements in approaches to predict variant effect/s. Such
advances will be critical to improve the sensitivity and specificity of
bioinformatic prediction of variant effect/s on SREs, BP sites, and
pseudoexon usage, and thereby improve assessment of variant
pathogenicity.