Bioinformatics Tool Review: Genome Variant Analysis Tools
Curator: Stephen J. Williams, Ph.D.
Updated 11/15/2018
The following post will be an ongoing curation of reviews of gene variant bioinformatic software.
The Ensembl Variant Effect Predictor.
McLaren W, Gil L, Hunt SE, Riat HS, Ritchie GR, Thormann A, Flicek P, Cunningham F.
Genome Biol. 2016 Jun 6;17(1):122. doi: 10.1186/s13059-016-0974-4.
Author information
1
European Molecular Biology Laboratory, European Bioinformatics Institute, Wellcome Genome Campus, Hinxton, Cambridge, CB10 1SD, UK. wm2@ebi.ac.uk.
2
European Molecular Biology Laboratory, European Bioinformatics Institute, Wellcome Genome Campus, Hinxton, Cambridge, CB10 1SD, UK.
3
European Molecular Biology Laboratory, European Bioinformatics Institute, Wellcome Genome Campus, Hinxton, Cambridge, CB10 1SD, UK. fiona@ebi.ac.uk.
Abstract
The Ensembl Variant Effect Predictor is a powerful toolset for the analysis, annotation, and prioritization of genomic variants in coding and non-coding regions. It provides access to an extensive collection of genomic annotation, with a variety of interfaces to suit different requirements, and simple options for configuring and extending analysis. It is open source, free to use, and supports full reproducibility of results. The Ensembl Variant Effect Predictor can simplify and accelerate variant interpretation in a wide range of study designs.
Rare diseases can be difficult to diagnose due to low incidence and incomplete penetrance of implicated alleles however variant analysis of whole genome sequencing can identify underlying genetic events responsible for the disease (Nature, 2015). However, a large cohort is required for many WGS association studies in order to produce enough statistical power for interpretation (see post and here). To this effect major sequencing projects have been initiated worldwide including:
- Iceland
- UK (100,000 Genome Project)
- USA
- Genome 10K
A more thorough curation of sequencing projects can be seen in the following post:
Icelandic Population Genomic Study Results by deCODE Genetics come to Fruition: Curation of Current genomic studies
And although sequencing costs have dramatically been reduced over the years, the costs to determine the functional consequences of such variants remains high, as thorough basic research studies must be conducted to validate the interpretation of variant data with respect to the underlying disease, as only a small fraction of variants from a genome sequencing project will encode for a functional protein. Correct annotation of sequences and variants, identification of correct corresponding reference genes or transcripts in GENCODE or RefSeq respectively offer compelling challenges to the proper identification of sequenced variants as potential functional variants.
To this effect, the authors developed the Ensembl Variant Effect Predictor (VEP), which is a software suite that performs annotations and analysis of most types of genomic variation in coding and non-coding regions of the genome.
Summary of Features
- Annotation: VEP can annotate two broad categories of genomic variants
- Sequence variants with specific and defined changes: indels, base substitutions, SNVs, tandem repeats
- Larger structural variants > 50 nucleotides
- Species and assembly/genomic database support: VEP can analyze data from any species with assembled genome sequence and annotated gene set. VEP supports chromosome assemblies such as the latest GRCh38, FASTA, as well as transcripts from RefSeq as well as user-derived sequences
- Transcript Annotation: VEP includes a wide variety of gene and transcript related information including NCBI Gene ID, Gene Symbol, Transcript ID, NCBI RefSeq ID, exon/intron information, and cross reference to other databases such as UniProt
- Protein Annotation: Protein-related fields include Protein ID, RefSeq ID, SwissProt, UniParc ID, reference codons and amino acids, SIFT pathogenicity score, protein domains
- Noncoding Annotation: VEP reports variants in noncoding regions including genomic regulatory regions, intronic regions, transcription binding motifs. Data from ENCODE, BLUEPRINT, and NIH Epigenetics RoadMap are used for primary annotation. Plugins to the Perl coding are also available to link other databases which annotate noncoding sequence features.
- Frequency, phenotype, and citation annotation: VEP searches Ensembl databases containing a large amount of germline variant information and checks variants against the dbSNP single nucleotide polymorphism database. VEP integrates with mutational databases such as COSMIC, the Human Gene Mutation Database, and structural and copy number variants from Database of Genomic Variants. Allele Frequencies are reported from 1000 Genomes and NHLBI and integrates with PubMed for literature annotation. Phenotype information is from OMIM, Orphanet, GWAS and clinical information of variants from ClinVar.
- Flexible Input and Output Formats: VEP supports input data format called “variant call format” or VCP, a standard in next-gen sequencing. VEP has the ability to process variant identifiers from other database formats. Output formats are tab deliminated and give the user choices in presentation of results (HTML or text based)
- Choice of user interface
- Online tool (VEP Web): simple point and click; incorporates Instant VEP Functionality and copy and paste features. Results can be stored online in cloud storage on Ensembl.
- VEP script: VEP is available as a downloadable PERL script (see below for link) and can process large amounts of data rapidly. This interface is powerfully flexible with the ability to integrate multiple plugins available from Ensembl and GitHub. The ability to alter the PERL code and add plugins and code functions allows the flexibility to modify any feature of VEP.
- VEP REST API: provides robust computational access to any programming language and returns basic variant annotation. Can make use of external plugins.
Watch Video on VES Instructional Webinar: https://youtu.be/7Fs7MHfXjWk
Watch Video on VES Web Version training on How to Analyze Your Sequence in VEP
Availability of data and materials
The dataset supporting the conclusions of this article is available from Illumina’s Platinum Genomes [93] and using the Ensembl release 75 gene set. Pre-built data sets are available for all Ensembl and Ensembl Genomes species [94]. They can also be downloaded automatically during set up whilst installing the VEP.
- Project name: Ensembl Variant Effect Predictor
- Project home page: http://www.ensembl.org/vep
- Archived version: https://github.com/Ensembl/ensembl-tools/archive/release/83.zip
- Zenodo deposit: https://zenodo.org/record/50492#.Vx9TJ5MrKEI
- Operating system: platform independent
- Programming language: Perl
- Other requirements: Perl 5.10 or higher and the DBI and DBD::mysql modules
- License: Apache 2.0
- Any restrictions to use by non-academics: none.
References
Large-scale discovery of novel genetic causes of developmental disorders.
Deciphering Developmental Disorders Study.
Nature. 2015 Mar 12;519(7542):223-8. doi: 10.1038/nature14135. PMID:25533962
Updated 11/15/2018
Research Points to Caution in Use of Variant Effect Prediction Bioinformatic Tools
Although we have the ability to use high throughput sequencing to identify allelic variants occurring in rare disease, correlation of these variants with the underlying disease is often difficult due to a few concerns:
- For rare sporadic diseases, classical gene/variant association studies have proven difficult to perform (Meyts et al. 2016)
- As Whole Exome Sequencing (WES) returns a considerable number of variants, how to differentiate the normal allelic variation found in the human population from disease-causing pathogenic alleles
- For rare diseases, pathogenic allele frequencies are generally low
Therefore, for these rare pathogenic alleles, the use of bioinformatics tools in order to predict the resulting changes in gene function may provide insight into disease etiology when validation of these allelic changes might be experimentally difficult.
In a 2017 Genes & Immunity paper, Line Lykke Andersen and Rune Hartmann tested the reliability of various bioinformatic software to predict the functional consequence of variants of six different genes involved in interferon induction and sixteen allelic variants of the IFNLR1 gene. These variants were found in cohorts of patients presenting with herpes simplex encephalitis (HSE). Most of the adult population is seropositive for Herpes Simplex Virus (HSV) however a minor fraction (1 in 250,000 individuals per year) of HSV infected individuals will develop HSE (Hjalmarsson et al., 2007). It has been suggested that HSE occurs in individuals with rare primary immunodeficiencies caused by gene defects affecting innate immunity through reduced production of interferons (IFN) (Zhang et al., Lim et al.).
References
Meyts I, Bosch B, Bolze A, Boisson B, Itan Y, Belkadi A, et al. Exome and genome sequencing for inborn errors of immunity. J Allergy Clin Immunol. 2016;138:957–69.
Hjalmarsson A, Blomqvist P, Skoldenberg B. Herpes simplex encephalitis in Sweden, 1990-2001: incidence, morbidity, and mortality. Clin Infect Dis. 2007;45:875–80.
Zhang SY, Jouanguy E, Ugolini S, Smahi A, Elain G, Romero P, et al. TLR3 deficiency in patients with herpes simplex encephalitis. Science. 2007;317:1522–7.
Lim HK, Seppanen M, Hautala T, Ciancanelli MJ, Itan Y, Lafaille FG, et al. TLR3 deficiency in herpes simplex encephalitis: high allelic heterogeneity and recurrence risk. Neurology. 2014;83:1888–97.
Genes Immun. 2017 Dec 4. doi: 10.1038/s41435-017-0002-z.
Frequently used bioinformatics tools overestimate the damaging effect of allelic variants.
Andersen LL1, Terczyńska-Dyla E1, Mørk N2, Scavenius C1, Enghild JJ1, Höning K3, Hornung V3,4, Christiansen M5,6, Mogensen TH2,6, Hartmann R7.
Abstract
We selected two sets of naturally occurring human missense allelic variants within innate immune genes. The first set represented eleven non-synonymous variants in six different genes involved in interferon (IFN) induction, present in a cohort of patients suffering from herpes simplex encephalitis (HSE) and the second set represented sixteen allelic variants of the IFNLR1 gene. We recreated the variants in vitro and tested their effect on protein function in a HEK293T cell based assay. We then used an array of 14 available bioinformatics tools to predict the effect of these variants upon protein function. To our surprise two of the most commonly used tools, CADD and SIFT, produced a high rate of false positives, whereas SNPs&GO exhibited the lowest rate of false positives in our test. As the problem in our test in general was false positive variants, inclusion of mutation significance cutoff (MSC) did not improve accuracy.
Methodology
- Identification of rare variants
- Genomes of nineteen Dutch patients with a history of HSE sequenced by WES and identification of novel HSE causing variants determined by filtering the single nucleotide polymorphisms (SNPs) that had a frequency below 1% in the NHBLI Exome Sequencing Project Exome Variant Server and the 1000 Genomes Project and were present within 204 genes involved in the immune response to HSV.
- Identified variants (204) manually evaluated for involvement of IFN induction based on IDBase and KEGG pathway database analysis.
- In-silico predictions: Variants classified by the in silico variant pathogenicity prediction programs: SIFT, Mutation Assessor, FATHMM, PROVEAN, SNAP2, PolyPhen2, PhD-SNP, SNP&GO, FATHMM-MKL, MutationTaster2, PredictSNP, Condel, MetaSNP, and CADD. Each program returned prediction scores measuring likelihood of a variant either being ‘deleterious’ or ‘neutral’. Prediction accuracy measured as
ACC = (true positive+true negative)/(true positive+true negative+false positive+false negative)
- Validation of prediction software/tools
In order to validate the predictive value of the software, HEK293T cells, deficient in IRF3, MAVS, and IKKe/TBK1, were cotransfected with the nine variants of the aforementioned genes and a luciferase reporter under control of the IFN-b promoter and luciferase activity measured as an indicator of IFN signaling function. Western blot was performed to confirm the expression of the constructs.
Results
| Table 2 Summary of the bioinformatic predictions |
HSE variants | IFNLR1 variants | Overall ACC | ||||||||||
| TN | TP | FN | FP | Total | ACC | TN | TP | FN | FP | Total | ACC | ||
| Uniform cutoff | |||||||||||||
| SIFT | 4 | 1 | 0 | 4 | 9 | 0.56 | 8 | 1 | 0 | 7 | 16 | 0.56 | 0.56 |
| Mutation assessor | 6 | 1 | 0 | 2 | 9 | 0.78 | 9 | 1 | 0 | 6 | 16 | 0.63 | 0.68 |
| FATHMM | 7 | 1 | 0 | 1 | 9 | 0.89 | 0.89 | ||||||
| PROVEAN | 8 | 1 | 0 | 0 | 9 | 1.00 | 11 | 1 | 0 | 4 | 16 | 0.75 | 0.84 |
| SNAP2 | 5 | 1 | 0 | 3 | 9 | 0.67 | 8 | 0 | 1 | 7 | 16 | 0.50 | 0.56 |
| PolyPhen2 | 6 | 1 | 0 | 2 | 9 | 0.78 | 12 | 1 | 0 | 3 | 16 | 0.81 | 0.80 |
| PhD-SNP | 7 | 1 | 0 | 1 | 9 | 0.89 | 11 | 1 | 0 | 4 | 16 | 0.75 | 0.80 |
| SNPs&GO | 8 | 1 | 0 | 0 | 9 | 1.00 | 14 | 1 | 0 | 1 | 16 | 0.94 | 0.96 |
| FATHMM MKL | 4 | 1 | 0 | 4 | 9 | 0.56 | 13 | 0 | 1 | 2 | 16 | 0.81 | 0.72 |
| MutationTaster2 | 4 | 0 | 1 | 4 | 9 | 0.44 | 14 | 0 | 1 | 1 | 16 | 0.88 | 0.72 |
| PredictSNP | 6 | 1 | 0 | 2 | 9 | 0.78 | 11 | 1 | 0 | 4 | 16 | 0.75 | 0.76 |
| Condel | 6 | 1 | 0 | 2 | 9 | 0.78 | 0.78 | ||||||
| Meta-SNP | 8 | 1 | 0 | 0 | 9 | 1.00 | 11 | 1 | 0 | 4 | 16 | 0.75 | 0.84 |
| CADD | 2 | 1 | 0 | 6 | 9 | 0.33 | 8 | 0 | 1 | 7 | 16 | 0.50 | 0.44 |
| MSC 95% cutoff | |||||||||||||
| SIFT | 5 | 1 | 0 | 3 | 9 | 0.67 | 8 | 1 | 0 | 8 | 16 | 0.50 | 0.56 |
| PolyPhen2 | 6 | 1 | 0 | 2 | 9 | 0.78 | 13 | 1 | 0 | 3 | 16 | 0.81 | 0.80 |
| CADD | 4 | 1 | 0 | 4 | 9 | 0.56 | 7 | 0 | 1 | 9 | 16 | 0.44 | 0.48 |
Note: TN: true negative, TP: true positive, FN: false negative, FP: false positive, ACC: accuracy
Functional testing (data obtained from reporter construct experiments) were considered as the correct outcome.
Three prediction tools (PROVEAN, SNP&GO, and MetaSNP correctly predicted the effect of all nine variants tested.
Other articles related to Genomics and Bioinformatics on this online Open Access Journal Include:
Finding the Genetic Links in Common Disease: Caveats of Whole Genome Sequencing Studies
Large-scale sequencing does not support the idea that lower-frequency variants have a major role in predisposition to type 2 diabetes
US Personalized Cancer Genome Sequencing Market Outlook 2018 –
Icelandic Population Genomic Study Results by deCODE Genetics come to Fruition: Curation of Current genomic studies
