Comparison of in Silico Strategies to Prioritize Rare Genomic Variants Impacting RNA Splicing for the Diagnosis of Genomic Disorders

Abstract

The development of computational methods to assess pathogenicity of pre-messenger RNA splicing variants is critical for diagnosis of human disease. We assessed the capability of eight algorithms, and a consensus approach, to prioritize 250 variants of uncertain significance (VUSs) that underwent splicing functional analyses. It is the capability of algorithms to differentiate VUSs away from the immediate splice site as being ‘pathogenic’ or ‘benign’ that is likely to have the most substantial impact on diagnostic testing. We show that SpliceAI is the best single strategy in this regard, but that combined usage of tools using a weighted approach can increase accuracy further. We incorporated prioritization strategies alongside diagnostic testing for rare disorders. We show that 15% of 2783 referred individuals carry rare variants expected to impact splicing that were not initially identified as ‘pathogenic’ or ‘likely pathogenic’; 1 in 5 of these cases could lead to new or refined diagnoses.

MRSD: A quantitative approach for assessing suitability of RNA-seq in the investigation of mis-splicing in Mendelian disease

“…This is amplified by unexpected impacts that many variants may have on mRNA splicing. 6 The MRSD-based approach that we describe here allows informed selection of biosample(s) for bulk RNA-seq, based on the required number of sequencing reads for appropriate surveillance of genes of interest. This enables effective patient-specific identification of genomic variants that are amenable for functional assessment of missplicing through RNA-seq.…”

“…Pathogenic variants, both protein-coding and intronic, that lie outside canonical splice sites may nonetheless act to disrupt pre-mRNA splicing through a diverse series of mechanisms (Figure S1). [1][2][3] Effective identification of pathogenic splice-impacting variants remains challenging and is limited by the omission of intronic regions in targeted sequencing approaches, 4,5 discordance between in silico variant prioritization tools, 6 and the lack of availability of the appropriate tissue from which to survey RNA for splicing disruption. 7,8 Targeted analyses such as RT-PCR enable detection of splicing aberrations 3 but are designed to test for the presence of specific disruptions.…”

A systematic analysis of splicing variants identifies new diagnoses in the 100,000 Genomes Project

“…Overall, we demonstrate the clinical value of examining both canonical and non-canonical splicing variants in unsolved rare diseases. 16…”

“…Recent advances in computation and artificial intelligence have led to the development of numerous in silico predictors for the prioritization of splicing variants 14 . For example, SpliceAI is a machine learning tool which robustly predicts splice sites and splice-disrupting variants 15 , and out-performs other algorithms in predicting splicing consequences from sequence data 16 . However, in clinical variant interpretation, well-validated functional assays have greater weight than in silico predictions of variant effect 6 , and functional validation of most splicing variants is still required to confirm a molecular diagnosis.…”

Splicing in the Diagnosis of Rare Disease: Advances and Challenges

“…Many of these tools perform well in the small number of studies that have compared their usage. In addition to Wai et al, Rowlands et al also found SpliceAI to be the single best predictive tool in their assessment of 9 tools’ performance on 250 VUS, although they did report greater accuracy from a weighted combination of multiple prediction tools, highlighting a strategy to improve predictions further, since combining multiple tools may mitigate individual weaknesses ( Rowlands et al, 2021a ). Surveying GT>GC splice donor variants, Chen et al tested SpliceAI’s ability to distinguish those which disrupt splicing with those that maintain normal splicing since these variants are not universally disruptive ( Chen et al, 2020 ).…”