In silico prediction of splice-altering single nucleotide variants in the human genome.
Jian X, Boerwinkle E, Liu X. · 2014
Read In silico prediction of splice-altering single nucleotide variants in the human genome. by Jian X, Boerwinkle E, Liu X. (2014) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4267638.txt.
Introduction
Xueqiu Jian, Eric Boerwinkle, and Xiaoming Liu test computer tools that predict whether a DNA change will disrupt how a gene's message is spliced. Their study appeared in Nucleic Acids Research in 2014. Many tools give scores that are hard to read, and few had been checked on a large shared set of variants.
The team compared eight tools on close to three thousand variants in splicing regions. Two older methods did best on their own. The team then built combined models that beat both and give scores a user can read at once. Applied to a cancer mutation catalog, the scores point to splicing changes as a real part of how some cancers start.
Catalog record
| CREATOR | Jian X, Boerwinkle E, Liu X. |
|---|---|
| YEAR | 2014 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
In silico prediction of splice-altering single nucleotide variants in the human genome By Xueqiu Jian, Eric Boerwinkle, Xiaoming Liu Published in: Nucleic Acids Research DOI: 10.1093/nar/gku1206 ABSTRACT Abstract In silico tools have been developed to predict variants that may have an impact on pre-mRNA splicing. The major limitation of the application of these tools to basic research and clinical practice is the difficulty in interpreting the output. Most tools only predict potential splice sites given a DNA sequence without measuring splicing signal changes caused by a variant. Another limitation is the lack of large-scale evaluation studies of these tools. We compared eight in silico tools on 2959 single nucleotide variants within splicing consensus regions (scSNVs) using receiver operating characteristic analysis. The Position Weight Matrix model and MaxEntScan outperformed other methods. Two ensemble learning methods, adaptive boosting and random forests, were used to construct models that take advantage of individual methods. Both models further improved prediction, with outputs of directly interpretable prediction scores. We applied our ensemble scores to scSNVs from the Catalogue of Somatic Mutations in Cancer database. Analysis showed that predicted splice-altering scSNVs are enriched in recurrent scSNVs and known cancer genes. We pre-computed our ensemble scores for all potential scSNVs across the human genome, providing a whole genome level resource for identifying splice-altering scSNVs discovered from large-scale sequencing studies. INTRODUCTION Since pre-mRNA splicing was first discovered in the 1970s (1,2), DNA variations that disrupt normal splicing have been linked to human genetic diseases (3,4,5). Unlike non-synonymous mutations within coding regions that directly alter amino acids by changing the codon, splice-altering mutations influence the normal process of removing introns from the pre-mRNA and rejoining the remaining exons. This normal process is regulated by complicated mechanisms that usually result in the production of different proteins by exon skipping, intron retention, use of different 5′ or 3′ splice sites, etc. which is termed alternative splicing (6). Alternative splicing is very common in human genes; it has been estimated to occur in ∼95% of multi-exon genes (7). The ubiquitous alternative splicing, the previous focus of the scientific community on exonic variants that directly modify protein sequences, and the fact that more than 20% of non-synonymous mutations reported in the Human Gene Mutation Database (HGMD) (8) may also affect splicing (9,10), create challenges in identifying variants that are causal or modifiers for human diseases due to disruption of splicing. Additionally, widespread high-throughput next-generation sequencing (NGS) technology is rapidly generating a large amount of data, which enables the identification of more variants in a shorter time than ever before. For example, a recent study sequenced the whole genome of 962 individuals and identified a total of more than 25 million genetic variants (11). This not only provides us with opportunities to discover novel causal variants but also makes the prioritization of these newly identified variants more challenging in view of the infeasibility of confirming each variant in vivo/in vitro. In silico tools can take advantage of sequence information to predict the possible effect of a variant based on specific models on a very large scale. This approach has been used to prioritize millions of variants (i.e. distinguish pathogenic mutations from a large number of background variations) and narrow down the search to a relatively small number of variants for laboratory validation. In particular, in silico tools for the prediction of variants affecting splicing have been developed that take into account different aspects of the splicing mechanism, which consist of (i) splicing signals, including the 5′/3′ splice site and the branch point; (ii) splicing regulatory elements, i.e. exonic/intronic splicing enhancers/silencers; and (iii) the spliceosome and other trans-acting elements that bind to cis-acting elements. Some tools have had successful application in predicting splicing defects. For instance, MaxEntScan, a tool that predicts the 5′/3′ splice site (12), correctly predicted three apparently nonsense, missense or silent mutations as disrupting normal splice sites in the ATM gene that are responsible for ataxia-telangiectasia (13). ESEfinder, a tool that predicts exonic splicing enhancers (ESEs) (14), has successfully predicted the loss of a putative ESE motif in the SMN2 gene as the cause of spinal muscular atrophy (15). These examples imply that in silico tools have the potential to be utilized for prioritization of variants that may disrupt splicing in the NGS era. Recently, we reviewed some of the most commonly used in silico tools for splicing defect prediction and noted that the major problem that prohibits the use of these tools in research and clinical practice is the difficulty in interpreting the output (16). One reason for this difficulty is that there is no unified standard to measure how splicing signals change when one allele is substituted by another because most tools only output prediction scores for potential splice sites given an input DNA sequence. Another reason is the lack of large-scale studies to evaluate the predictive performance of these tools. To make a more thorough comparison of existing prediction tools and provide a directly interpretable score for splice-altering variants, we evaluated eight in silico tools and constructed prediction models using ensemble learning methods that take advantage of some of these tools in order to further improve prediction. Because the relatively conserved sequences at the two ends of an intron (splicing consensus sequences) are the prominent cis-acting elements of splicing (17) that have been the focus of most existing tools, and single nucleotide variants (SNVs) are the most commonly observed variants in the human genome (18), the present study is restricted to SNVs within splicing consensus regions (−3 to +8 at the 5′ splice site and −12 to +2 at the 3′ splice site) defined by Burge et al. (17); we named these variants scSNVs. We validated our models, pre-computed our ensemble prediction scores for all potential scSNVs across the human genome and applied them to scSNVs from the Catalogue of Somatic Mutations in Cancer (COSMIC) database (19). MATERIALS AND METHODS Data sources Positive (i.e. splice-altering) variants were downloaded from three databases: (i) the HGMD Professional Version 2013.1, which contains more than 13 000 mutations with consequences for mRNA splicing (8); most (more than 8000) are located at invariant GT-AG sites (the first two and last two sites of an intron), while the remaining sites are mostly exonic; (ii) the SpliceDisease database, which collects and curates experimentally supported data of RNA splicing mutations and disease (20); this database integrated 2337 splicing mutation-disease entries, including 303 genes and 370 human diseases from 898 publications; and (iii) the Database for Aberrant Splice Sites (DBASS), which contains 577 and 307 records of mutation-induced and disease-causing aberrant 5′ and 3′ splice sites, respectively (21). Negative variants were retrieved from the 1000 Genomes Project phase 1 data (18), which contain genomes of 1092 individuals from 14 populations that can be used as controls for comparisons with other samples. To further validate our final models, an additional independent test set was retrieved from the published data of Houdayer et al., in which the impact of the variants was experimentally validated (22). Splicing variant filtering strategy The following strategies were used when recruiting variants from the databases into our positive group: (i) based on the National Center for Biotechnology Information Reference Sequence (RefSeq) database release 59 (23), we