snpTree - a web-server to identify and construct SNP trees from whole genome sequence data
© Leekitcharoenphon et al.; licensee BioMed Central Ltd. 2012
Published: 13 December 2012
The advances and decreasing economical cost of whole genome sequencing (WGS), will soon make this technology available for routine infectious disease epidemiology. In epidemiological studies, outbreak isolates have very little diversity and require extensive genomic analysis to differentiate and classify isolates. One of the successfully and broadly used methods is analysis of single nucletide polymorphisms (SNPs). Currently, there are different tools and methods to identify SNPs including various options and cut-off values. Furthermore, all current methods require bioinformatic skills. Thus, we lack a standard and simple automatic tool to determine SNPs and construct phylogenetic tree from WGS data.
Here we introduce snpTree, a server for online-automatic SNPs analysis. This tool is composed of different SNPs analysis suites, perl and python scripts. snpTree can identify SNPs and construct phylogenetic trees from WGS as well as from assembled genomes or contigs. WGS data in fastq format are aligned to reference genomes by BWA while contigs in fasta format are processed by Nucmer. SNPs are concatenated based on position on reference genome and a tree is constructed from concatenated SNPs using FastTree and a perl script. The online server was implemented by HTML, Java and python script.
The server was evaluated using four published bacterial WGS data sets (V. cholerae, S. aureus CC398, S. Typhimurium and M. tuberculosis). The evalution results for the first three cases was consistent and concordant for both raw reads and assembled genomes. In the latter case the original publication involved extensive filtering of SNPs, which could not be repeated using snpTree.
The snpTree server is an easy to use option for rapid standardised and automatic SNP analysis in epidemiological studies also for users with limited bioinformatic experience. The web server is freely accessible at http://www.cbs.dtu.dk/services/snpTree-1.0/.
The dramatic decrease in cost for whole-genome sequencing (WGS) has made this technology economically feasible as a routine tool for scientific research, including infectious disease epidemiology. In addition, WGS has major applications for health service providers working with infectious diseases  as such to deliver high-resolution genomic epidemiology as the ultimate typing method for bacteria.
The ideal microbial typing technique should enable differentiation of epidemiological unrelated strains and group epidemiological related (outbreak) strains,  and give information that will help to understand the evolutionary history of multiple strains within a clonal lineage [1, 2]. Although some current technologies are highly informative like MLST or PFGE, they have limited resolution when applied to closely related isolates and different methods often have to be applied in different situations [1, 2].
Especially outbreak isolates normally have very little diversity and require extensive genomic methods to differentiate and catagorize the isolates . Single nucleotide polymorphisms (SNPs) also show relatively low mutation rates and are evolutionarily stable. Moreover, SNPs analysis has successfully been used for determining broad patterns of evolution in many recent studies [4–6].
Currently, There are a number of available non-commercial NGS genotype analysis software such as SOAP2 , GATK  and SAMtools . Nonetheless, all of the software require bioinformatic skills, various options, various setting and they do not have a user friendly web-interface.
Here we introduce snpTree. A server for online-automatic SNP analysis and SNP tree construction from sequencing reads as well as from assembled genomes or contigs. The server is a pipeline which intregrates avaliable SNPs analysis softwares such as SAMtools  and MUMmer , with customized scripts. The performance of the server was evaluated with four published bacterial WGS data set; Vibrio cholerae , Staphylococcus aureus CC398 , Salmonella Typhimurium  and Mycobacterium tuberculosis .
The web-server contains more than 2,000 completed reference genomes collected from NCBI Genome database (accessed on April 2012).
SNPs identification from WGS
Prior to mapping raw reads to a proper reference genome, the sequence data in fastq format are filtered and trimmed according to the following criteria : (i) reads with N's are removed, (ii) if a read matches a minimum of 25 nt of a sequencing primer/adaptor the reads are trimmed at the 5' coordinate of match, (iii) the 3' tail bases are trimmed if the quality score is less than 20, (iv) the minimum average quality of the read should be 20 and the read length after trimming should be at least 20 nt.
Trimmed raw reads are aligned against a reference genome using BWA  with minimum mapping quality equal to 30 as a default (Figure 1A). BWA is based on an effective data compression algorithm called Burrows-Wheeler transform (BWT) that is fast, memory-efficient and espectially useful for aligning short reads .
SNPs calling and filtering are accomplished by SAMtools that is a software package for parsing and manipulating alignments in the generic alignment format (SAM/BAM format) . The snpTree server allows users to set a couple of parameters to filter SNPs, a minimum coverage and a minimum distance between each SNPs (prune). The default for both cut-offs is set to 10 and additionally all heterozygous SNPs are filtered because these are likely mapping errors in haploid chromosomes. The identifed SNPs are concluded into a VCF file.
SNPs identification from assembled genomes
A pipeline has been developed around the software package MUMmer version 3.23  (Figure 1B). An application named Nucmer, which is part of MUMmer, is used to align each of de novo assemblies to a reference genome chosen by the user (default settings). SNPs are then called from the resulting alignments with another MUMmer application named "show-snps" (with options "-CIlrT"). A pruning is then applied, if chosen by the user, and the SNPs are written into a VCF formatted file for each of the analyzed genomes.
SNPs tree construction
One VCF formatted file is needed for each Operational Taxonomic Unit (OTU). The SNPs are then concatenated into a single alignment by ignoring indels. Including indels would disturb the position of SNPs in the sigle alignment. To include indels in any trees, it requires some sensible way to represent them numerically as distances in an evolutionary space, and there is no any ways to achieve this. Indels could theoretically be included in a multiple sequence alignment, since such alignments can handle gaps but it's difficult to score them. "Blast-like" gap penalties certainly would not work, since they are optimized for much larger gaps, e.g. recombination events.
It is important to note that SNPs not found in a VCF file is interpreted as not being a variation and the corresponding base in the reference is expected. This might not always be the right choice, because a SNP not found in a VCF file could be a result of an INDEL. It is expected to be a rare case and probably won't disturb the phylogenetic signal.
The alignment is passed on to Fastree , which creates a maximum likelihood tree from the SNP alignment.
snpTree server output
Results and discussion
The snpTree was evaluated using raw reads and assembled genomes from four published bacterial WGS data sets (V. cholerae , S. aureus CC398 , S. Typhimurium  and M. tuberculosis ). The evaluation was considered based on tree topology as well as the reference genome's position of identifed SNPs.
Evaluation of tree topology and SNPs position
WGS from published data set were subjected to snpTree server in order to generate SNP trees. The tree topology evaluation was based on percentage of concordance. If the strain in the tree from snpTree server matches exactly with the tree from published data, it was considered as an exact match. If the strains were grouped into the same cluster with published data, it was considered as a cluster match. In addition, the snpTree server was evaluated with assembled genomes or contigs. The raw reads were assembled prior by de novo assembly using Velvet 1.1.04 . The assembled genomes were processed to snpTree server to make SNP trees.
V. cholerae data set
Percentage of concordance
V. cholerae (raw reads)
V. cholerae (contigs)
S. aureus CC398 (raw reads)
S. aureus CC398 (contigs)
S. typhimurium (raw reads)
S . typhimurium (contigs)
M. tuberculosis (raw reads)
M. tuberculosis (contigs)
S.aureus CC398 data set
For S. aureus CC398 (Table 1), snpTree produced a tree with 87 - 88 % concordance for exact match and 96 - 97 % concordance for cluster match. SNP trees for raw reads and assembled genomes are shown in Additional file 3 and Additional file 4 respectively. There were 91 and 90 % overlapping SNPs for raw reads and assembled genomes (Figure 4). The performance of snpTree on this data set was slightly less than for the V. cholera data set. The reason is probably that the genomes of 89 S. aureus CC398 isolates came from animals and humans sources from 19 countries and four continents. In addition, there are 4,238 SNPs among them . These isolates are more diverse than V. cholera isolates. Thus, this diversity makes difficulty for snpTree to capture exactly the same variant as in original publication. Nevertheless, snpTree can differentiate between isolates from humans and pigs which is very meaningful to epidemiological studies.
S. Typhimurium data set
The third data set, S. Typhimurium, which consists of 51 Salmonella in which 43 isolates from 14 patients with multiple recurrences in Blantyre, Malawi and 8 control typhimurium isolates . Like in the original publication, both raw reads and contigs data set, the isolates fell within three distint phylogenetic clusters (Additional file 5 and 6) which gave 100 % concordance for cluster match (Table 1). On the other hand, the percentage of concordance for exact match was quite low (53 - 61 %). It is not possible to evaluate SNPs position for this data set because of lacking SNPs position data. However, the number of identified SNPs from snpTree server (1,692 SNPs) was not much different from original data set (1,463 SNPs). Most of the S. Typhimurium isolates are highly genetically related as they came from patients who had recrudescence and/or reinfections. Therefore, this study requires high-resolution SNPs analysis and intensive phylogenetic tree construction to differentiate these little variation. In addition, the original tree from this data set was generated and confirmed using several independent approaches, with bootstrap support and clade credibility marked  which snpTree cannot repeat as using bootstrapping is time-consuming.
M.tuberculosis data set
Another data set that consists of 32 M. tuberculosis outbreak isolates and 4 historical isolates (from the same region but isolated before the outbreak) with matching genotype suggesting that the outbreak was clonal . The performance of snpTree server on this data set was inconsistent due to low concordance percentage for exact match and cluster match (Table 1, Additional file 7 and 8). Moreover, the number of indentified SNPs and matching SNP positions (Figure 3) are very different between the tree from snpTree server (677 SNPs) and the published data (204 SNPs). The original publication determined transmission dynamics of the outbreak at a higher resolution by filtering to remove many of SNPs in repetitive regions and those appearing in a single isolate. Thus, the procedure in the original manuscript is impossible to repeat and it should be noted that the original filtering reduced the number of SNP's from more than 1,000 to 204. This is probably the reason that snpTree were unable to reproduce the same results as in the original publication.
Sensitivity and specificity
Sensitivity and specificity
Variable and cut-off value
Number of bp between SNPs
Number of bp from a sequence end
The sensitivity for prune cut-off (Table 2) was slightly dropped when increasing number of prune. This is due to the more number of bp between SNPs (prune) leading to the high chance to have SNPs between that number of bp.
Using minimum number of bp from a sequence end as a varied cut-off, the sensitivity was very high and stable for all varied values. It is quite rare to have SNPs occurred in the tails of sequence so this cut-off less affects to the SNP calling process. The specificity for both cut-off were very high. It is because the number of SNP inserted is extreamly low (1,000 SNPs) compared to the whole genome (4,878,012 bp).
The rapid technological advantages in WGS and rapidly decreasing cost has made the technology available for large groups of scientists as well as clinical microbiologists. It is expected that WGS will very soon find widespread use in clinical and public health microbiology, as has already been shown . The implementation of such technologies will however, create a major need for simple to use bioinformatic tools to make sense of the data generated. We have here developed snpTree and evaluated it on four different published datasets. The concordance of the SNPs tree from raw reads was more adequate than the one from assembled genomes, which is not surprising. However, in practice transfering sequencing reads will be more time-consuming than just transferring assembled genomes and the tree topology from these different kind of genomes was only sligthly different. Therefore, the assembled genomes option in snpTree server can provide a quicker solution for uploading time-consuming. In order to create informative SNPs tree, using a closely related reference genome is important. Therefore, the selection of a proper reference genome is crucial. Thus, it is adviced to choose a reference genome belonging to the same or as closely related a sub-type as possible to the strain collection under study. This could for species where this is a available reference belonging to the same MLST type. In the future a more generic solution to overcome this obstracle might be to using high-resolution prediction method such as K-mers to assign a genuine reference genome.
The advance of WGS and the use of epidemiological genomics underline the potential of practical application of WGS for clinical microbiology and emphazies the importance of biology and evolution in developing reliable and accurate genomics tools for clinical use. In addition, SNP-typing phylogenetic methods can distinguish very closely related isolates to a degree not achievable by widely employed sub-genomic typing tools. snpTree server might be not a perfect tool but it is an option for easy and rapid standardised and automatic SNP analysis tool in epidemiological studies. It is also useful for users with limited bioinformatic experience.
This study was supported by the Center for Genomic Epidemiology (09- 067103/DSF) http://www.genomicepidemiology.org and Danish Food Industry Agency (3304-FVFP-08). PL and RKM would like to acknowledge funding from the Technical University of Denmark.
This article has been published as part of BMC Genomics Volume 13 Supplement 7, 2012: Eleventh International Conference on Bioinformatics (InCoB2012): Computational Biology. The full contents of the supplement are available online at http://www.biomedcentral.com/bmcgenomics/supplements/13/S7.
- Parkhill J, Wren BW: Bacterial epidemiology and biology--lessons from genome sequencing. Genome biology. 2011, 12: 230-10.1186/gb-2011-12-10-230.PubMed CentralView ArticlePubMedGoogle Scholar
- Foxman B, Zhang L, Koopman JS, Manning SD, Marrs CF: Choosing an appropriate bacterial typing technique for epidemiologic studies. Epidemiologic perspectives & innovations. 2005, 2: 10-10.1186/1742-5573-2-10.View ArticleGoogle Scholar
- Hendriksen RS, Price LB, Schupp JM, Gillece JD, Kaas RS, Engelthaler DM, Bortolaia V, Pearson T, Waters AE, Upadhyay BP, Shrestha SD, Adhikari S, Shakya G, Keim PS, Aarestrup FM: Population genetics of Vibrio cholerae from Nepal in 2010: evidence on the origin of the Haitian outbreak. MBio. 2011, 2:Google Scholar
- Harris SR, Feil EJ, Holden MT, Quail MA, Nickerson EK, Chantratita N, Gardete S, Tavares A, Day N, Lindsay JA, Edgeworth JD, de Lencastre H, Parkhill J, Peacock SJ, Bentley SD: Evolution of MRSA during hospital transmission and intercontinental spread. Science. 2010, 327: 469-74. 10.1126/science.1182395.PubMed CentralView ArticlePubMedGoogle Scholar
- Grad YH, Lipsitch M, Feldgarden M, Arachchi HM, Cerqueira GC, Fitzgerald M, Godfrey P, Haas BJ, Murphy CI, Russ C, Sykes S, Walker BJ, Wortman JR, Young S, Zeng Q, Abouelleil A, Bochicchio J, Chauvin S, Desmet T, Gujja S, McCowan C, Montmayeur A, Steelman S, Frimodt-Møller J, Petersen AM, Struve C, Krogfelt KA, Bingen E, Weill FX, Lander ES, Nusbaum C, Birren BW, Hung DT, Hanage WP: Genomic epidemiology of the Escherichia coli O104:H4 outbreaks in Europe, 2011. Proceedings of the National Academy of Sciences of the United States of America. 2012, 109: 3065-70. 10.1073/pnas.1121491109.PubMed CentralView ArticlePubMedGoogle Scholar
- Price LB, Stegger M, Hasman H, Aziz M, Larsen J, Andersen PS, Pearson T, Waters AE, Foster JT, Schupp J, Gillece J, Driebe E, Liu CM, Springer B, Zdovc I, Battisti A, Franco A, Zmudzki J, Schwarz S, Butaye P, Jouy E, Pomba C, Porrero MC, Ruimy R, Smith TC, Robinson DA, Weese JS, Arriola CS, Yu F, Laurent F, Keim P, Skov R AF: Staphylococcus aureus CC398: Host Adaptation and Emergence of Methicillin Resistance in Livestock. MBio. 2012, 3: 1-6. 10.3391/mbi.2012.3.1.01.View ArticleGoogle Scholar
- Li R, Li Y, Fang X, Yang H, Wang J, Kristiansen K, Wang J: SNP detection for massively parallel whole-genome resequencing. Genome Res. 2009, 19: 1124-1132. 10.1101/gr.088013.108.PubMed CentralView ArticlePubMedGoogle Scholar
- McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, Garimella K, Altshuler D, Gabriel S, Daly M, DePristo MA: The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome research. 2010, 20: 1297-303. 10.1101/gr.107524.110.PubMed CentralView ArticlePubMedGoogle Scholar
- Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G, Durbin R, 1000 Genome Project Data Processing Subgroup: The Sequence Alignment/Map format and SAMtools. Bioinformatics. 2009, 25: 2078-9. 10.1093/bioinformatics/btp352.PubMed CentralView ArticlePubMedGoogle Scholar
- Delcher AL, Phillippy A, Carlton J, Salzberg SL: Fast algorithms for large-scale genome alignment and comparison. Nucleic acids research. 2002, 30: 2478-83. 10.1093/nar/30.11.2478.PubMed CentralView ArticlePubMedGoogle Scholar
- Okoro CK, Kingsley RA, Quail MA, Kankwatira AM, Feasey NA, Parkhill J, Dougan G, Gordon MA: High-resolution single nucleotide polymorphism analysis distinguishes recrudescence and reinfection in recurrent invasive nontyphoidal salmonella typhimurium disease. Clinical infectious diseases. 2012, 54: 955-63. 10.1093/cid/cir1032.PubMed CentralView ArticlePubMedGoogle Scholar
- Gardy JL, Johnston JC, Ho Sui SJ, Cook VJ, Shah L, Brodkin E, Rempel S, Moore R, Zhao Y, Holt R, Varhol R, Birol I, Lem M, Sharma MK, Elwood K, Jones SJ, Brinkman FS, Brunham RC, Tang P: Whole-genome sequencing and social-network analysis of a tuberculosis outbreak. The New England journal of medicine. 2011, 364: 730-9. 10.1056/NEJMoa1003176.View ArticlePubMedGoogle Scholar
- Li H, Durbin R: Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics. 2009, 25: 1754-60. 10.1093/bioinformatics/btp324.PubMed CentralView ArticlePubMedGoogle Scholar
- Larsen MV, Cosentino S, Rasmussen S, Friis C, Hasman H, Marvig RL, Jelsbak L, Sicheritz-Pontén T, Ussery DW, Aarestrup FM, Lund O: Multilocus Sequence Typing of Total Genome Sequenced Bacteria. Journal of clinical microbiology. 2012, 1355-1361.Google Scholar
- Nielsen R, Paul JS, Albrechtsen A, Song YS: Genotype and SNP calling from next-generation sequencing data. Nature reviews. Genetics. 2011, 12: 443-51. 10.1038/nrg2986.PubMed CentralView ArticlePubMedGoogle Scholar
- Price MN, Dehal PS, Arkin AP: FastTree: computing large minimum evolution trees with profiles instead of a distance matrix. Molecular biology and evolution. 2009, 26: 1641-50. 10.1093/molbev/msp077.PubMed CentralView ArticlePubMedGoogle Scholar
- Danecek P, Auton A, Abecasis G, Albers CA, Banks E, DePristo MA, Handsaker RE, Lunter G, Marth GT, Sherry ST, McVean G, Durbin R, 1000 Genomes Project Analysis Group: The variant call format and VCFtools. Bioinformatics. 2011, 27: 2156-8. 10.1093/bioinformatics/btr330.PubMed CentralView ArticlePubMedGoogle Scholar
- Zerbino DR, Birney E: Velvet: algorithms for de novo short read assembly using de Bruijn graphs. Genome research. 2008, 18: 821-9. 10.1101/gr.074492.107.PubMed CentralView ArticlePubMedGoogle Scholar
- Eyre DW, Golubchik T, Gordon NC, Bowden R, Piazza P, Batty EM, Ip CL, Wilson DJ, Didelot X, O'Connor L, Lay R, Buck D, Kearns AM, Shaw A, Paul J, Wilcox MH, Donnelly PJ, Peto TE, Walker AS, Crook DW: A pilot study of rapid benchtop sequencing of Staphylococcus aureus and Clostridium difficile for outbreak detection and surveillance. BMJ Open. 2012, 2:Google Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.