GTB – an online genome tolerance browser
© The Author(s). 2017
Received: 18 March 2016
Accepted: 17 December 2016
Published: 6 January 2017
Accurate methods capable of predicting the impact of single nucleotide variants (SNVs) are assuming ever increasing importance. There exists a plethora of in silico algorithms designed to help identify and prioritize SNVs across the human genome for further investigation. However, no tool exists to visualize the predicted tolerance of the genome to mutation, or the similarities between these methods.
We present the Genome Tolerance Browser (GTB, http://gtb.biocompute.org.uk): an online genome browser for visualizing the predicted tolerance of the genome to mutation. The server summarizes several in silico prediction algorithms and conservation scores: including 13 genome-wide prediction algorithms and conservation scores, 12 non-synonymous prediction algorithms and four cancer-specific algorithms.
The GTB enables users to visualize the similarities and differences between several prediction algorithms and to upload their own data as additional tracks; thereby facilitating the rapid identification of potential regions of interest.
KeywordsSNVs Mutation Pathogenicity prediction Prediction algorithm Variant effect prediction Genome browser Genome tolerance
The rate at which single nucleotide variants (SNVs) are being identified across the genome has increased owing to technological advances and the falling costs in whole-genome sequencing . The main challenge facing clinicians and researchers is identifying which of these SNVs contribute to disease predisposition . There are many algorithms capable of predicting the functional consequences of these variants, including those focussing on nonsynonymous SNVs (nsSNVs) that induce amino acid substitutions [4, 18], SNVs that influence specific diseases such as cancer [7, 17], or SNVs that fall within non-coding regions of the genome [8, 14, 19]. However, each method employs a different approach to variant effect prediction, which can sometimes lead to conflicting predictions for the same variant being made. For example, sequence-based algorithms begin with a multiple sequence alignment between the gene or protein of interest and homologous sequences. Here, it is assumed that conserved positions within the alignment indicate that there are strong selective pressures acting on particular residues; therefore, genomic variants occurring at these positions are often considered to be functional. On the other hand, structure-based algorithms use structural properties, such as the accessible solvent area, to identify putative functional variants. These algorithms assume that variants falling at specific sites are functional regardless of sequence conservation, e.g. buried residues. Recently, a new class of prediction algorithms capitalizing on state-of-the-art machine learning paradigms have emerged. These algorithms combine several sequence and structure-based annotations to train classifiers using known disease-causing variants and neutral polymorphisms. A comprehensive review on the underlying methodology of prediction algorithms is given in Ng and Henikoff , and a comprehensive comparative evaluation of algorithm performance has been performed by Thusberg et al .
The wealth of available prediction algorithms makes assessing the predicted impact of genomic variants a tedious and time consuming task. As a result, databases such as the dbNSFP  and the dbWGFP  have begun to collate the output of several different prediction algorithms; thereby allowing users to assess the concordance between prediction algorithms. While the reported correlation between existing algorithms varies considerably, ranging from near zero to near perfect correlation , no tool exists for visualizing these similarities and differences. In this work, we present the Genome Tolerance Browser (GTB): an online browser for visualizing the predicted tolerance of the genome to mutation and for identifying potential similarities and subtle differences between in silico prediction algorithms.
Construction and content
Prediction algorithms and conservation scores
List of in silico prediction algorithms and conservation scores summarized through the Genome Tolerance Browser
Non-Synonymous Prediction Algorithms
PolyPhen-2 (HumVar & HumDiv)
FATHMM (Unweighted & Weighted)
Genome-Wide Prediction Algorithms
PhyloP (46-way; vertebrate, primates and placental mammals)
PhyloP (100-way; vertebrate, primates and placental mammals)
Calculating genome tolerance
A web-based version of the GTB is available at http://gtb.biocompute.org.uk and has been built on top of the Dalliance genome browser . By default, tracks representing two popular non-synonymous prediction algorithms: SIFT and PolyPhen-2, and two genome-wide prediction algorithms: FATHMM-MKL and CADD, are displayed. Using the available options, users can add additional tracks representing a plethora of computational prediction algorithms (see Table 1 for a full list of available methods), or even upload custom annotation data in either bigWig or bigBed format. The appearance of these tracks can be customized, and publication quality images can be exported in either SVG or PNG format. Users can also download the entire GTB database or extract GTB scores for specific regions by following the instructions given on the website.
In the following section, we demonstrate how the GTB can be used to visualize, compare and contrast several prediction algorithms. Understanding why various algorithms agree in particular regions, but disagree in other regions, is an important aspect when interpreting computational predictions. In addition, when multiple algorithms all yield different predictions and/or tolerance profiles, this could suggest that variants falling in these regions are much harder to predict. Therefore, users should treat predictions with caution and not rely on a single algorithm for interpretation. Further, the browser can also be used to identify potential “regions of interest”. Here, long stretches of intolerance predicted by multiple algorithms may indicate regions worth exploring through in vitro experimentation.
Visualizing the characteristics of sequence- and structure-based prediction algorithms
Although SIFT shows higher intolerance across HOXA5, the overall profile shows similar regions of intolerance to that of PROVEAN. For example, both appear to show high intolerance towards the end of the 1st exon (see region highlighted in red). However, this comes as no surprise given that these genes play a crucial role during embryonic development and are highly conserved across great evolutionary distances . In contrast, PolyPhen-2, which incorporates structure-based properties for variant prioritization, shows a different tolerance profile. Here, it appears that it is specific regions of HOX5A that are intolerant to mutation. This suggests that these regions may harbour important structural constraints which are potentially missed when using a pure sequence-based approach. Both PolyPhen-2 models, HumVar and HumDiv, share large regions of similarity (highlighted in red). However, this also comes as no surprise as they both utilize the same underlying prediction algorithm but are trained using slightly different training data . Peaks of predicted intolerance can also be observed across the non-coding region of HOXA5 when using genome-wide prediction algorithms such as FATHMM-MKL and CADD; thereby suggesting that these regions could also be functional. However, it is interesting to note that FATHMM-MKL appears to give much more granular peaks across the region than CADD. Both algorithms are trained using similar genomic annotations. Therefore, this observation appears to suggest that these algorithms may place greater emphasis on different genomic annotations across HOXA5.
Visualizing the impact of cancer-specific training
The Genome Tolerance Browser (GTB) offers a platform to effectively compare and visualize differences in functional predictions between a wide range of algorithms at (or below) the gene level. This enables the researcher to clearly understand the nature of differences in performance and make a more informed decision about the best algorithm to use for a particular scenario. For example, the browser can be used to identify cases in which particular algorithms place greater emphasis on similar annotations during prediction, as illustrated by the emphasis on sequence conservation we observed when comparing SIFT and PROVEAN. The GTB can also be used to detect subtle differences between prediction algorithms. For example, we observed clear discrepancies in predicted intolerance between generic prediction algorithms and cancer-specific prediction algorithms across cancer-associated regions of the genome, illustrating that these different methodologies place greater emphasis on different annotations during prediction.
The potential utility of the GTB goes beyond simply visualizing computational prediction algorithms. For example, other research questions that could be asked include: are prediction algorithms affected by genomic annotations such as open chromatin, transcription factor binding sites and histone modifications; and can some of the observed variability between prediction algorithms be explained by these annotations; given specific genomic annotations, under what circumstances should we use particular prediction algorithms (or particular methodologies towards prediction)?
Finally, the GTB can be used to identify potential regions of interest across the genome, e.g. long stretches of predicted intolerance. In future releases, we plan on developing algorithms for automatically detecting and characterizing these regions of interest.
The GTB is a visualization platform that enables users to compare a range of existing variant effect prediction algorithms (and other data as additional tracks) in specific regions of the human genome. The GTB enables differences in prediction to be evaluated and facilitates rapid identification of potential regions of interest.
Availability and requirements
Genome tolerance browser
non-synonymous single nucleotide variant
Portable network graphics
Single nucleotide variant
Scalable vector graphics
This work was supported by the Medical Research Council (MC UU 12013/8 and G1000427/1). M.R. was supported by an EPSRC grant (EP/K008250/1).
Developed the software: HAS. Wrote the manuscript: HAS. Tested the software, contributed ideas to development, edited and commented on manuscript: MFR, MF, CC, TRG. Acquired funding: CC, TRG. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Consent for publication
Ethics approval and consent to participate
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Adzhubei IA, Schmidt S, Peshkin L, Ramensky VE, Gerasimova A, Bork P, Kondrashov AS, Sunyaev SR. A method and server for predicting damaging missense mutations. Nat Methods. 2010;7:248–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Choi Y, Sims GE, Murphy S, Miller JR, Chan AP. Predicting the Functional Effect of Amino Acid Substitutions and Indels. PLoS ONE. 2012;7:e46688.View ArticlePubMedPubMed CentralGoogle Scholar
- Davydov EV, Goode DL, Sirota M, Cooper GM, Sidow A, Batzoglou S. Identifying a High Fraction of the Human Genome to be under Selective Constraint Using GERP++. PLoS Comput Biol. 2010;6:e1001025.View ArticlePubMedPubMed CentralGoogle Scholar
- Dong C, Wei P, Jian X, Gibbs R, Boerwinkle E, Wang K, Liu X. Comparison and integration of deleteriousness prediction methods for nonsynonymous SNVs in whole exome sequencing studies. Hum Mol Genet. 2015;24:2125–37.View ArticlePubMedGoogle Scholar
- Down TA, Piipari M, Hubbard TJP. Dalliance: interactive genome viewing on the web. Bioinforma (Oxford, England). 2011;27:889–90.View ArticleGoogle Scholar
- Erzurumluoglu AM, Rodriguez S, Shihab HA, Baird D, Richardson TG, Day INM, Gaunt TR. Identifying Highly Penetrant Disease Causal Mutations Using Next Generation Sequencing: Guide to Whole Process. Biomed Res Int. 2015;2015:e923491.View ArticleGoogle Scholar
- Gonzalez-Perez A, Deu-Pons J, Lopez-Bigas N. Improving the prediction of the functional impact of cancer mutations by baseline tolerance transformation. Genome Med. 2012;4:89.View ArticlePubMedPubMed CentralGoogle Scholar
- Kircher M, Witten DM, Jain P, O’Roak BJ, Cooper GM, Shendure J. A general framework for estimating the relative pathogenicity of human genetic variants. Nat Genet. 2014;46:310–5.View ArticlePubMedPubMed CentralGoogle Scholar
- Liu X, Jian X, Boerwinkle E. dbNSFP: a lightweight database of human nonsynonymous SNPs and their functional predictions. Hum Mutat. 2011;32:894–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Liu X, Wu C, Li C, Boerwinkle E. dbNSFP v3.0: A One-Stop Database of Functional Predictions and Annotations for Human Nonsynonymous and Splice-Site SNVs. Hum Mutat. 2016;37:235–41.View ArticlePubMedGoogle Scholar
- Ng PC, Henikoff S. Predicting deleterious amino acid substitutions. Genome Res. 2001;11:863–74.View ArticlePubMedPubMed CentralGoogle Scholar
- Ng PC, Henikoff S. Predicting the effects of amino acid substitutions on protein function. Annu Rev Genomics Hum Genet. 2006;7:61–80.View ArticlePubMedGoogle Scholar
- Olivier M, Hollstein M, Hainaut P. TP53 Mutations in Human Cancers: Origins, Consequences, and Clinical Use. Cold Spring Harb Perspect Biol. 2010;2:a001008.View ArticlePubMedPubMed CentralGoogle Scholar
- Quang D, Chen Y, Xie X. DANN: a deep learning approach for annotating the pathogenicity of genetic variants. Bioinformatics. 2015;31:761–3.View ArticlePubMedGoogle Scholar
- Reva B, Antipin Y, Sander C. Predicting the functional impact of protein mutations: application to cancer genomics. Nucleic Acids Res. 2011;39:e118.View ArticlePubMedPubMed CentralGoogle Scholar
- Santini S, Boore JL, Meyer A. Evolutionary Conservation of Regulatory Elements in Vertebrate Hox Gene Clusters. Genome Res. 2003;13:1111–22.View ArticlePubMedPubMed CentralGoogle Scholar
- Shihab HA, Gough J, Cooper DN, Day INM, Gaunt TR. Predicting the functional consequences of cancer-associated amino acid substitutions. Bioinformatics. 2013;29:1504–10.View ArticlePubMedPubMed CentralGoogle Scholar
- Shihab HA, Gough J, Cooper DN, Stenson PD, Barker GLA, Edwards KJ, Day INM, Gaunt TR. Predicting the Functional, Molecular, and Phenotypic Consequences of Amino Acid Substitutions using Hidden Markov Models. Hum Mutat. 2013;34:57–65.View ArticlePubMedGoogle Scholar
- Shihab HA, Rogers MF, Gough J, Mort M, Cooper DN, Day INM, Gaunt TR, Campbell C. An integrative approach to predicting the functional effects of non-coding and coding sequence variation. Bioinformatics. 2015;31(10):1536-43.
- Soutar AK, Naoumova RP. Mechanisms of Disease: genetic causes of familial hypercholesterolemia. Nat Clin Pract Cardiovasc Med. 2007;4:214–25.View ArticlePubMedGoogle Scholar
- The 1000 Genomes Project Consortium. An integrated map of genetic variation from 1,092 human genomes. Nature. 2012;491:56–65.View ArticlePubMed CentralGoogle Scholar
- Thusberg J, Olatubosun A, Vihinen M. Performance of mutation pathogenicity prediction methods on missense variants. Hum Mutat. 2011;32:358–68.View ArticlePubMedGoogle Scholar
- Welcsh PL, King MC. BRCA1 and BRCA2 and the genetics of breast and ovarian cancer. Hum Mol Genet. 2001;7:705–13.View ArticleGoogle Scholar
- Wu J, Wu M, Li L, Liu Z, Zeng W, Jiang R. dbWGFP: a database and web server of human whole-genome single nucleotide variants and their functional predictions. Database: The Journal of Biological Databases and Curation. 2016. baw024. http://doi.org/10.1093/database/baw024.