MIST: A Tool for Rapid in silico Generation of Molecular Data from

Bacterial Genome Sequences

Peter Kruczkiewicz

1,2

, Steven Mutschall

, Dillon Barker

1,5

, James Thomas

, Gary Van Domselaar

Victor P. J. Gannon

, Catherine D. Carrillo

and Eduardo N. Taboada

1,2

Laboratory for Foodborne Zoonoses, Public Health Agency of Canada, Lethbridge, AB, Canada

Department of Biological Sciences, University of Lethbridge, Lethbridge, AB, Canada

Bureau of Microbial Hazards, Food Directorate, Health Canada, Ottawa, ON, Canada

National Microbiology Laboratory, Public Health Agency of Canada, Winnipeg, MB, Canada

Department of Microbiology, University of Manitoba, Winnipeg, MB, Canada

Keywords:

Molecular Typing, MLST, CGF, PCR, MLVA, VNTR.

Abstract:

Whole-genome sequence (WGS) data can, in principle, resolve bacterial isolates that differ by a single base

pair, thus providing the highest level of discriminatory power for epidemiologic subtyping. Nonetheless, be-

cause the capability to perform whole-genome sequencing in the context of epidemiological investigations

involving priority pathogens has only recently become practical, fewer isolates have WGS data available rela-

tive to traditional subtyping methods. It will be important to link these WGS data to data in traditional typing

databases such as PulseNet and PubMLST in order to place them into proper historical and epidemiological

context, thus enhancing investigative capabilities in response to public health events. We present MIST (Mi-

crobial In Silico Typer), a bioinformatics tool for rapidly generating in silico typing data (e.g. MLST, MLVA)

from draft bacterial genome assemblies. MIST is highly customizable, allowing the analysis of existing typing

methods along with novel typing schemes. Rapid in silico typing provides a link between historical typing

data and WGS data, while also providing a framework for the assessment of molecular typing methods based

on WGS analysis.

1 INTRODUCTION

A key component of the public health response to pri-

ority pathogens involves efforts to implement broad-

based systems for epidemiologic surveillance. These

activities are aimed at examining geographical/host

distribution and sources of disease, with the ultimate

goal of reducing the burden of illness through targeted

interventions that disrupt pathogen transmission path-

ways.

The last two decades have witnessed the emer-

gence of DNA-based subtyping methods, which have

become integral to epidemiology, making it one the

most powerful tools in public health (van Belkum,

2003). Genotyping methods have revolutionized the

ﬁeld of epidemiology by enhancing the ability to dis-

criminate between isolates, which has made it invalu-

able in a range of contexts: early outbreak detection;

microbial source attribution to determine sources of

contamination and exposure; identiﬁcation of the

types and subtypes of pathogenic bacteria circulating

among humans, animals, and the environment (Fox-

man and Riley, 2001). Various subtyping schemes

have been developed based on the analysis of differ-

ent types of polymorphisms that can be broadly clas-

siﬁed into three major groups: DNA banding-pattern

based methods, such as pulsed-ﬁeld gel electrophore-

sis (PFGE) and restriction fragment length polymor-

phism (RFLP); DNA hybridization-based methods,

such as PCR and microarrays; and DNA sequence-

based methods, such as multilocus sequence typing

(MLST) (Li et al., 2009). Although these methods

have been applied with varying levels of success, the

increasing resolution of modern methods has greatly

enhanced epidemiologic investigations. For example,

by identifying cases with matching subtyping data, it

has been possible to distinguish outbreak cases from

316

Krukczkiewicz P., Mutschall S., Barker D., Thomas J., Van Domselaar G., P. J. Gannon V., D. Carrillo C. and N. Taboada E..

MIST: A Tool for Rapid in silico Generation of Molecular Data from Bacterial Genome Sequences.

DOI: 10.5220/0004249003160323

In Proceedings of the International Conference on Bioinformatics Models, Methods and Algorithms (BIOINFORMATICS-2013), pages 316-323

ISBN: 978-989-8565-35-8

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

concurrent sporadic cases (Singh et al., 2006).

While a few leading techniques remain widely

used tools for strain characterization, high-throughput

whole-genome sequencing (WGS) has rapidly be-

come a viable option for the investigation of prior-

ity pathogens in the context of public health events

(Gilmour et al., 2010; Alexander et al., 2012). In

principle, WGS data can resolve bacterial isolates that

differ by a single base pair, providing the highest level

of discrimination for bacterial strain characterization.

However, because the capability to perform WGS for

subtyping has only recently become practical, rela-

tively few isolates have WGS data available compared

to traditional typing methods such as PFGE or MLST,

which have dedicated databases with extensive histor-

ical data going back a decade or more. Thus, a crit-

ical component in the transition from molecular epi-

demiology based on existing genotyping approaches

to a genomic epidemiology paradigm based on the

analysis of WGS data in an epidemiologic framework

will be the linkage of WGS data to databases such as

PulseNet (Swaminathan et al., 2001) and PubMLST

(Jolley et al., 2004). This will allow genomic data to

be placed into proper historical and epidemiological

context, thus enhancing the analysis of surveillance

data and the investigation of outbreak data.

This study presents Microbial In Silico Typer

(MIST), a tool for the rapid in silico generation of

molecular typing proﬁles (e.g. MLST, MLVA) from

draft bacterial genome sequences. We have recently

reported results obtained using MIST as part of a

framework for assessing the performance of existing

molecular typing methods for Campylobacter jejuni

and C. coli using WGS data (Carrillo et al., 2012).

In this study we present the full implementation of

MIST and demonstrate the utility of MIST analysis

using ﬁnished, closed and draft assembly genome se-

quence data from priority human pathogens, and show

through comparison with published and experimen-

tal data that MIST produces accurate in silico-derived

molecular subtyping data for a variety of existing and

novel typing approaches.

2 IMPLEMENTATION

2.1 Overview

MIST is a standalone application written in the C#

programming language using the Microsoft .NET

Framework 4.0. MIST uses ObjectListView 2.5.0,

an open-source extension to the .NET ListView,

and DotNetZip 1.9.1.8, an open-source library for

creating and extracting ZIP ﬁles. The GUI version

WGS assembly

Microbial In Silico Typer

Probe hy-

bridization

PCR

Sequence

typing

SNP

typing

VNTR

Sequence

type

Binary

ﬁngerprint

Hybrid-

ization

proﬁle

SNP des-

ignation

VNTR

proﬁle

In silico typing results

Figure 1: Overall MIST in silico typing workﬂow.

of MIST is currently only supported under Windows

(XP, Vista, 7). However, a command-line version

of MIST is available for both Windows and Linux.

MIST is a freely available application licensed under

the GNU General Public License. The application,

source code and documentation is available at

https://bitbucket.org/peterk87/microbialinsilicotyper

and https://bitbucket.org/peterk87/mist for the GUI

and command-line versions of MIST, respectively.

All analyses performed within MIST utilize se-

quence similarity searching, which is performed

using blastn from the NCBI BLAST+ application

suite (Altschul et al., 1990; Camacho et al., 2009).

MIST can produce in silico data for PCR-, probe

hybridization-, sequence-, variable number of tandem

repeat- (VNTR), and SNP-based typing assays (Fig-

ure 1) while also providing users with detailed sum-

mary data underlying the subtyping results.

MIST was designed such that the user is free to

use existing typing schemes or to create their own.

Novel schemes can be based on conventional sub-

typing approaches or can incorporate molecular tar-

gets that would be impractical in the context of a

laboratory-based assay. For a given bacterial species,

an existing Package, consisting of one or more typing

schemes or Assays, themselves comprised of one or

more Markers, can be loaded and executed in MIST,

yielding in silico results for the assays selected. An

assay creation module allows the user to easily edit or

create new packages and assays through an entry form

or through the import of a tab-delimited ﬁle with the

necessary assay parameters.

MIST allows the user to easily export typing re-

sults or to save and continue an analysis with addi-

tional WGS data or typing schemes.

MIST:AToolforRapidinsilicoGenerationofMolecularDatafromBacterialGenomeSequences

317

2.2 PCR-based Typing

WGS assembly

BLAST search

PCR primers

Primer binding sites

Find most likely amplicon

Correct

amplicon

size?

Present Absent

Determine copy number

Yes No

VNTR-based typing

Figure 2: In silico VNTR- and PCR-based typing workﬂow.

MIST simulates in silico PCR using BLAST se-

quence similarity searching (Figure 2). BLAST

searches are conducted to ﬁnd all potential for-

ward and reverse primer binding sites in the subject

genome. All combinations of potential forward and

reverse primer sites are investigated to determine pu-

tative amplicons. A positive result is reported when

an amplicon can be retrieved within the expected size

tolerances and in which the amplicon is ﬂanked by

the forward and reverse primers in either correct ori-

entation. If all primer binding site combinations have

been exhausted and no suitable amplicon is gener-

ated, a negative result is reported. Although tradi-

tional in vitro PCR assays produce only a binary sig-

nal, MIST provides a more in-depth analysis of PCR

results. A summary of binary absence-presence re-

sults along with detailed match information can be

accessed for each PCR marker match.

2.3 VNTR-based Typing

MIST simulates in silico VNTR results (i.e. MLVA)

by using a variation on in silico PCR analysis (Fig-

ure 2). If an amplicon can be retrieved with a length

less than the user-deﬁned maximum amplicon size

threshold (3000 bp by default), the copy number or

the number of repeats is reported. If a suitable am-

plicon cannot be found, a null result is returned. Al-

though the copy number reported is typically zero or

an integer, in the case of partial repeats a non-integer

number may be reported. To aid in the intrepretation

of the data, detailed match information, such as the

most likely primer binding sites and predicted ampli-

con, is made available.

2.4 Probe Hybridization-based Typing

WGS assembly

BLAST search

Probe

BLAST hit

above the

IDT?

AbsentPresent

Determine SNP designation

Yes No

SNP-based typing

Figure 3: In silico probe hybridization- and SNP-based typ-

ing workﬂow.

Two modes for in silico probe hybridization anal-

ysis are available in MIST. ‘Mode 1’, which is the

default setting, is for ‘short’ probes (i.e. < 70 bp or

user-deﬁned), such as those which might be found on

oligonucleotide-based microarrays. ‘Mode 2’ is for

‘long’ probes (i.e. > 70 bp or user-deﬁned), such as

those which might be found in amplicon-based mi-

croarrays. For ‘short’ probes the hybridization proﬁle

for a genome is predicted using BLAST based on a

user-deﬁned percent identity threshold (IDT) for each

probe in the assay. For ‘long’ probes, an alignment

length threshold (ALT) is used in addition to the IDT.

In both modes, BLAST is used to identify hits for

each probe; the hit with highest bit score is used if

multiple BLAST hits can be identiﬁed. In Mode 1,

a positive hit is reported if the top BLAST hit has a

percent identity equal to or greater than the IDT over

the entire length of the probe. If no BLAST hits can

be identiﬁed or if the top hit does not fulﬁll the IDT

criterion, a negative result is reported. In Mode 2, a

positive hit is reported if the top BLAST hit has a per-

cent identity and alignment length equal to or greater

than the IDT and ALT, respectively. If no BLAST hits

can be identiﬁed or if the top hit does not fulﬁll the

IDT and ALT criteria, a negative result is reported. In

addition to the binary absence-presence results MIST

provides detailed match information, which is also

BIOINFORMATICS2013-InternationalConferenceonBioinformaticsModels,MethodsandAlgorithms

318

provided for negative results, thus allowing the user

to verify results obtained.

2.5 SNP-based Typing

SNP loci can be interrogated using a variation of

the in silico probe hybridization (Figure 3). To de-

sign a SNP-based assay, the SNP position is indi-

cated by an ‘N’ within a probe sequence. Probe se-

quences are queried against the subject genome using

BLAST. The nucleotide at a SNP position is reported

if a match above the IDT can be found in the sub-

ject genome. Other match information made avail-

able includes the probe match sequence and the loca-

tion within the subject genome. If a match above the

IDT cannot be found within the subject genome, the

top BLAST hit is provided along with the number of

mismatches with respect to the query probe sequence.

2.6 Sequence-based Typing

WGS assembly

BLAST search

Allele sequences

Perfect

match

found?

Allele match Closest allele match

Yes No

Figure 4: In silico sequence-based typing workﬂow.

The allelic proﬁle of a genome for a given gene

marker is determined through sequence similarity

searching of a database of the known alleles for the

gene using BLAST (Figure 4). MIST utilizes a rapid

search strategy wherein one allele of each unique size

for each allele database is queried against the subject

genome. A simple matching algorithm is then used to

ﬁnd a perfect match in the allelic database. If a perfect

match cannot be found using this approach, complete

pairwise BLAST searching against the allele database

is performed to determine the closest matching alle-

les (i.e. the alleles with least number of mismatches)

along with the number of SNPs with respect to the

subject genome’s allele.

In addition to the individual allelic proﬁles, MIST

automatically determines the overall sequence type

(ST) and clonal complex (CC). If additional metadata

is linked to allelic proﬁles, this ancillary data can be

displayed along with ST and CC information.

3 RESULTS AND DISCUSSION

Wherever possible, MIST-derived typing results were

validated by comparison against experimentally-

derived typing results obtained from the literature.

The accuracy of MIST is inherently linked to the

quality of the WGS sequence data (i.e. the subject

genome) used as an input because genome assemblies

with large numbers of contigs, or regions of poor or

absent sequence data may introduce false negatives

to the in silico results (Carrillo et al., 2012). Simi-

larly, accuracy can be impacted by the error rate in

experimental data, which is largely due the inherent

noise in hybridization, a process upon which most

subtyping methods rely. Nevertheless, MIST was

found to generate sensitive analyses and in silico re-

sults that were highly concordant with experimental

data whether from completed whole genome data or

from minimally processed draft genome assemblies.

3.1 PCR-based Typing: Campylobacter

jejuni CGF40 Assay

The comparative genomic ﬁngerprinting (CGF) assay,

CGF40 (Taboada et al., 2012), was used to assess the

performance of in silico PCR analysis using MIST.

The CGF40 assay targets 40 accessory genes present

within C. jejuni NCTC 11168 and variably present in

other C. jejuni and C. coli strains. The 40 primer pairs

along with the expected amplicon sizes and an ampli-

con size tolerance of 5% were input into MIST as an

Assay.

Based on MIST CGF40 analysis of 25 sequenced

C. jejuni genomes, the average concordance between

in silico and in vitro-derived results was found to be

92.8% (928/1000 matches) using a blastn word size

of 8. The false negative and false positive rates were

4.0% and 3.2%, respectively. Generally, false nega-

tive results were attributable to sequence truncations

within incomplete draft assemblies. False positive

results were largely “true” amplicons that contained

SNPs within priming sites, which would likely affect

ampliﬁcation under standard PCR conditions. For ex-

ample, marker cj1721 yielded a concordance of only

48%, with 13 false positive discrepancies. A multi-

ple sequence alignment of the retrieved cj1721 am-

plicons revealed several SNPs in both priming sites.

However, after reducing the stringency of the test by

increasing the blastn word size, the false positive rate

for cj1721 dropped from 48% to 12% without ap-

preciably changing the overall concordance of the re-

maining markers in the assay.

Thus, the blastn word size can be used to adjust

the stringency of an in silico PCR test. A larger word

MIST:AToolforRapidinsilicoGenerationofMolecularDatafromBacterialGenomeSequences

319

size (e.g. >7) will produce results that are highly

speciﬁc (and less sensitive) with respect to the given

primer inputs; likely producing results that are more

concordant with in vitro PCR due to fewer false pos-

itive matches. Alternatively, a lower word size (e.g.

<7) will produce results that are more sensitive (and

less speciﬁc), which may be useful for exploratory

analyses where multiple sequence alignments of the

retrieved amplicons can be investigated to further ex-

amine differences between strains.

MIST offers user control over the in silico results

and provides a variety of output data making it a valu-

able tool in the validation and creation of novel PCR

typing schemes.

3.2 VNTR-based Typing: Listeria

monocytogenes MLVA

VNTR typing was simulated using an MLVA scheme

for L. monocytogenes (Miya et al., 2008) on 23 pub-

licly available L. monocytogenes genomes. In silico

VNTR data generated by MIST are shown in Table 1.

The in silico VNTR results for F2365 matched the ex-

pected results with 14, 17 and 9 repeats in the VNTR

loci TR1, TR2 and TR3, respectively.

It was not possible to validate results obtained

by MIST in any L. monocytogenes strains other than

F2365 because of the lack of available MLVA data on

genome-sequenced strains in the literature for this or

other MLVA schemes. MIST was unable to ﬁnd an

amplicon in some cases for the MLVA marker TR1.

This is consistent with previous observations on the

potential unsuitability of the TR1 locus due to vari-

ability between strains in that region (Miya et al.,

2008). Our results thus illustrate how MIST may be

a valuable tool in the in silico validation and develop-

ment of novel VNTR-based typing schemes.

3.3 Probe Hybridization-based Typing:

Escherichia coli Pathotyping

Microarray

Probe hybridization was simulated using the probe se-

quences from the E. coli miniaturized pathotyping mi-

croarray (Anjum et al., 2007) on 110 E. coli genomes

obtained from NCBI. The microarray consists of a

combination of 61 probes targeting virulence, bacte-

riocin and conserved genes that are shared among or

speciﬁc to E. coli pathotypes. Concordance was cal-

culated between the in silico and in vitro array for the

seven E. coli strains shared between the public and

study datasets.

The overall concordance for these seven E. coli

Table 1: The number of predicted repeats for the in silico

L. monocytogenes MLVA scheme (Miya et al., 2008) are

shown along with the lineage of each strain. In silico MLVA

data for F2365 is in boldface. In some cases with TR1,

MIST was unable to ﬁnd a suitable amplicon (missing data

indicated with (–)).

Strain Lineage TR1 TR2 TR3

F2365 I 14 17 9

J1-194 I 6 18 5

N1-017 I 15 46.83 6

R2-503 I 13 46.83 6

Clip80459 I 15 19 5

H7858 I 23 22 5

HPB2262 I – 12 7

Scott-A I 14 17 9

F6854 II 18.22 8.33 9

F6900 II 21 9.5 9

J0161 II 15.56 7.83 9

J2818 II 21 9.5 9

N3-165 II 16.11 8.33 9

1988 II – 8.33 9

10403S II 7.33 8 9

EGD-e II 26.44 11.5 9

R2-561 II 17.67 8 9

08-5923 II 35 12 9

08-5578 II 35 12 9

HCC23 III – 22 5

L99-4a III – 22 5

M7 III – 22 5

J2-071 III – 16 5

strains was 93.4% (399/427 matches) using a probe

percent identity threshold of 90%. The false nega-

tive and false positive rates were 2.8% and 3.8%, re-

spectively. The rrl 23S rRNA control gene probes

accounted for the majority of discrepancies (71.4%),

since four probes from this locus had mismatches in

more than half of the strains tested. Removal of the

rrl control gene probes increased the concordance to

97.9% (322/329 matches). Generally, false negative

results by MIST were attributable to SNPs within the

probe hybridization site resulting in a %ID below the

90% ID threshold. False positive results by MIST

were conﬁrmed to be full length or near full length

matches with greater than 95%ID (0-2 mismatches).

In silico pathotype designations are shown in Table 2.

Table 2: The in silico E. coli pathotyping microarray

analysis-derived pathotypes for 7 E. coli strains.

Sample Pathotype

536 ExPEC

IHE3034 ExPEC

K-12 MG1655 –

B171 EPEC;EHEC

CFT073 EAEC;ETEC;ExPEC

O127:H6 E2348-69 EAEC;ETEC;EPEC;EHEC

O157:H7 EDL933 EAEC;ETEC;EPEC;EHEC;STEC

BIOINFORMATICS2013-InternationalConferenceonBioinformaticsModels,MethodsandAlgorithms

320

3.4 SNP-based Typing: Listeria

monocytogenes TMLGT

An in silico SNP-based typing assay was created

based on the targeted multilocus genotyping assay

(TMLGT) for L. monocytogenes (Ward et al., 2010).

The 30 probe assay simultaneously classiﬁes the 4

lineages, 4 major serotypes and four epidemic clones

(EC) of L. monocytogenes by interrogating a 3’ SNP

diagnostic for one of the aforementioned groups. The

MIST assay for TMLGT was tested against 23 pub-

licly available L. monocytogenes genome sequences

and validated by comparing known lineage, serotype

and EC designations to the MIST-derived designa-

tions. MIST results for the TMLGT assay placed all

23 strains into their correct lineages. Serotypes were

correctly assigned to 15 out of 17 strains belonging to

one of the four serotypes targeted by this assay. All

eight strains belonging to known EC groups were cor-

rectly identiﬁed (Table 3).

SNP-based typing assays are particularly

amenable to analysis in MIST because the results

require little analytical interpretation, as a SNP vari-

ant determination is directly linked to the underlying

sequence data. Furthermore, because the assay is

performed in silico, literally any number of variant

sites can be tested for, allowing for the development

and testing of assays consisting of hundreds if not

thousands of markers, which would be difﬁcult to

deploy in a laboratory setting.

3.5 Sequence-based Typing:

Staphylococcus aureus MLST

Sequence-based typing was simulated using the

MLST scheme for Staphylococcus aureus (Enright

et al., 2000). Genome sequences for 63 publicly

available S. aureus strains for which correspond-

ing STs were available from the MLST database

(saureus.mlst.net) were examined. The concordance,

calculated by computing the number of matching al-

leles between in vitro and in silico MLST data as a

proportion of the total number of alleles, was found

to be 98.6% (435/441 matching alleles).

Six allelic discrepancies were found among ﬁve

genomes, and were investigated using the detailed

match information available within MIST. Five dis-

crepancies were cases in which a different allele num-

ber was assigned based on a minor SNP variant; four

cases differed by a single SNP, and one case differed

by two SNPs. In each case, results obtained with

MIST reﬂected a canonical match between the un-

derlying WGS data provided and the discordant al-

lele in the database. For the remaining case of allelic

Table 3: The in silico TMLGT analysis-derived lineage,

serotype and EC designations of 23 L. monocytogenes

strains are shown here. In silico-derived data not matching

in vitro-derived data are shown in brackets.

Sample Lineage Serotype EC

F2365 I 4b I

J1-194 I 1/2b –

N1-017 I (1/2b) –

R2-503 I 1/2b –

Clip80459 I 4b –

H7858 I 4b II

HPB2262 I 4b Ia

Scott-A I 4b Ia

F6854 II 1/2a III

F6900 II 1/2a III

J0161 II 1/2a III

J2818 II 1/2a III

N3-165 II 1/2a –

1988 II 1/2a –

10403S II 1/2a –

EGD-e II (1/2c) –

R2-561 II 1/2c –

08-5923 II 1/2a –

08-5578 II 1/2a –

HCC23 III – –

L99-4a III – –

M7 III – –

J2-071 III – –

discrepancy, the allele for arcC in genome TCH60

could not be identiﬁed. Further examination revealed

a single-base adenine deletion within a 7-base poly-

adenine tract; an error common in 454 pyrosequenc-

ing. After correcting for this assumed error, the

TCH60 arcC allele was correctly assigned by MIST

as arcC 2. These data not only show that MIST can

accurately determine MLST data with a very high de-

gree of conﬁdence, but also illustrate how the detailed

information provided by MIST can prove useful.

The MIST platform could be used in the testing

and development of expanded MLST schemes includ-

ing whole-genome MLST analysis, which are of great

interest in the context of rapid whole genome-based

subtyping of strains. Moreover, the provision of ad-

ditional subtype-associated metadata is a feature not

common among other in silico MLST tools, includ-

ing those provided by online MLST databases.

4 CONCLUSIONS

The current generation of molecular typing meth-

ods was originally conceived as a proxy for whole-

genome sequence data at a time when whole genome

sequencing had not yet been achieved, or more re-

cently, when the cost and logistics precluded the se-

MIST:AToolforRapidinsilicoGenerationofMolecularDatafromBacterialGenomeSequences

321

quencing of more than one or a small number of

strains for a particular species.

Ultimately, high quality, whole genome sequenc-

ing is the true gold standard for molecular character-

ization of microbes. In addition to encoding all of

the information necessary to determine conventional

molecular typing proﬁles, WGS provides the most

comprehensive data for inferring strain to strain phy-

logenetic relationships. For these reasons, WGS anal-

ysis has become increasingly prominent in the public

health response to pathogens, enabling characteriza-

tion of organisms at an unprecedented level of reso-

lution. Moreover, as the cost of sequencing continues

to decline, the applicability of WGS will broaden to

a larger scope of strains and, in the near future, will

likely become the method of choice for characteriza-

tion of all microbes.

In the context of public health, a signiﬁcant gap

currently exists in that although WGS analysis has

been shown to be viable in the context of public health

events such as outbreaks (Gilmour et al., 2010), it may

not yet be possible to perform WGS in the context of

routine epidemiologic surveillance. In silico typing

provides the opportunity for the utilization of WGS

data as a framework for comparing the performance

of existing molecular typing methods or for assessing

and validating novel methods (Carrillo et al., 2012).

Moreover, while ever-expanding, WGS datasets

still comprise only a fraction of the historical data

contained in public repositories of molecular typing

data, such as PulseNet (Swaminathan et al., 2001) and

PubMLST (Jolley et al., 2004), and this is not likely

to change in the immediate future. Establishing links

between WGS data and the data contained in molecu-

lar epidemiology databases will be critical as we tran-

sition from a paradigm involving molecular typing to

one that is based on WGS.

Recently developed comparative genomics-based

tools have been developed with the aforementioned

goals in mind. A web-based tool was recently devel-

oped for in silico MLST analysis of bacterial isolates

using WGS data (Larsen et al., 2012). Bacterial Iso-

late Genomic Sequence Database (BIGSDB), which

is based on the mlstdbNet platform (Jolley et al.,

2004), has been developed for storage and analysis

of WGS data along with other types of data such as

MLST data (Jolley and Maiden, 2010). BIGSDB has

been used in conjunction with Ion Torrent sequenc-

ing to rapidly generate accurate typing data on multi-

ple methods for Neisseria meningitidis (Vogel et al.,

2012). MIST was designed to provide a complimen-

tary approach to existing tools by allowing users the

ﬂexibility to analyze draft WGS data and produce in

silico typing proﬁles for traditional typing schemes,

assays based on modiﬁcations on these traditional

schemes, and assays based on novel typing schemes.

In silico typing proﬁles can be linked to histor-

ical typing data from public databases and to other

available metadata on each genome analysed, such

as phenotypic, clinical, or epidemiological informa-

tion, enabling the user to examine associations be-

tween the sub-typing data and their underlying meta-

data. Since MIST allows the user to design and test

schemes based on approaches that may be imprac-

tical to perform in vitro, extended MLST schemes

or hybridization-based methods targeting hundreds of

loci can be tested that would otherwise be subject

to sensitivity/speciﬁcity issues when deployed in the

lab. Furthermore, the MIST platform enables compar-

ison between typing methods through the simultane-

ous generation of in silico typing results for a variety

of different methods.

As we move into the era of broad-based WGS,

it is important to consider the relationship between

WGS data and historical molecular typing databases,

which are rich sources of molecular and epidemiolog-

ical information. Ideally, these sources of information

should be used cooperatively; large molecular epi-

demiological databases should be used as resources

for targeted selection of strains for WGS projects,

and the corollary, as WGS begins to replace to tra-

ditional genotyping, the ability to rapidly assess rel-

evant genetic markers from an abundance of mini-

mally processed sequence data will become an es-

sential requirement towards leveraging these data for

maximum public health impact. Continued efforts to-

ward the development of bioinformatics tools that en-

able rapid analysis of draft genome data will enable

public health laboratories to not only link WGS data

to historical data but provide a framework for use in

the development of novel analytical methods and pro-

vide an improved understanding of the performance

of current methods to assist in the interpretation of

existing studies.

REFERENCES

Alexander, D. C., Hao, W., Gilmour, M. W., Zittermann,

S., Sarabia, A., Melano, R. G., Peralta, A., Lombos,

M., Warren, K., Amatnieks, Y., Virey, E., Ma, J. H.,

Jamieson, F. B., Low, D. E., and Allen, V. G. (2012).

Escherichia coli O104:H4 infections and international

travel. Emerging Infectious Diseases, 18(3):473–476.

PMID: 22377016.

Altschul, S. F., Gish, W., Miller, W., Myers, E. W., and

Lipman, D. J. (1990). Basic local alignment search

tool. Journal of Molecular Biology, 215(3):403–410.

Anjum, M. F., Mafura, M., Slickers, P., Ballmer, K., Kuh-

BIOINFORMATICS2013-InternationalConferenceonBioinformaticsModels,MethodsandAlgorithms

322

nert, P., Woodward, M. J., and Ehricht, R. (2007).

Pathotyping Escherichia coli by using miniaturized

DNA microarrays. Applied and Environmental Mi-

crobiology, 73(17):5692–5697.

Camacho, C., Coulouris, G., Avagyan, V., Ma, N., Pa-

padopoulos, J., Bealer, K., and Madden, T. L. (2009).

BLAST+: architecture and applications. BMC Bioin-

formatics, 10:421. PMID: 20003500.

Carrillo, C. D., Kruczkiewicz, P., Mutschall, S., Tudor, A.,

Clark, C., and Taboada, E. N. (2012). A framework for

assessing the concordance of molecular typing meth-

ods and the true strain phylogeny of Campylobac-

ter jejuni and C. coli using draft genome sequence

data. Frontiers in Cellular and Infection Microbiol-

ogy, 2:57.

Enright, M. C., Day, N. P. J., Davies, C. E., Peacock, S. J.,

and Spratt, B. G. (2000). Multilocus sequence typ-

ing for characterization of methicillin-resistant and

methicillin-susceptible clones of Staphylococcus au-

reus. Journal of Clinical Microbiology, 38(3):1008–

1015.

Foxman, B. and Riley, L. (2001). Molecular epidemiology:

focus on infection. American Journal of Epidemiol-

ogy, 153(12):1135–1141. PMID: 11415945.

Gilmour, M. W., Graham, M., Domselaar, G. V., Tyler,

S., Kent, H., Trout-Yakel, K. M., Larios, O., Allen,

V., Lee, B., and Nadon, C. (2010). High-throughput

genome sequencing of two Listeria monocytogenes

clinical isolates during a large foodborne outbreak.

BMC Genomics, 11(1):120.

Jolley, K. A., Chan, M., and Maiden, M. C. (2004). mlstdb-

Net - distributed multi-locus sequence typing (MLST)

databases. BMC Bioinformatics, 5:86–86. PMID:

15230973 PMCID: 459212.

Jolley, K. A. and Maiden, M. C. J. (2010). BIGSdb: scal-

able analysis of bacterial genome variation at the pop-

ulation level. BMC Bioinformatics, 11:595. PMID:

21143983.

Larsen, M. V., Cosentino, S., Rasmussen, S., Friis, C., Has-

man, H., Marvig, R. L., Jelsbak, L., Sicheritz-Pont

en,

T., Ussery, D. W., Aarestrup, F. M., and Lund, O.

(2012). Multilocus sequence typing of total-genome-

sequenced bacteria. Journal of Clinical Microbiology,

50(4):1355–1361. PMID: 22238442.

Li, W., Raoult, D., and Fournier, P. (2009). Bacterial strain

typing in the genomic era. FEMS Microbiology Re-

views, 33(5):892–916. PMID: 19453749.

Miya, S., Kimura, B., Sato, M., Takahashi, H., Ishikawa, T.,

Suda, T., Takakura, C., Fujii, T., and Wiedmann, M.

(2008). Development of a multilocus variable-number

of tandem repeat typing method for Listeria mono-

cytogenes serotype 4b strains. International Jour-

nal of Food Microbiology, 124(3):239–249. PMID:

18457891.

Singh, A., Goering, R. V., Simjee, S., Foley, S. L., and

Zervos, M. J. (2006). Application of molecular tech-

niques to the study of hospital infection. Clinical Mi-

crobiology Reviews, 19(3):512–530.

Swaminathan, B., Barrett, T. J., Hunter, S. B., and Tauxe,

R. V. (2001). PulseNet: the molecular subtyp-

ing network for foodborne bacterial disease surveil-

lance, united states. Emerging Infectious Diseases,

7(3):382–389. PMID: 11384513.

Taboada, E. N., Ross, S. L., Mutschall, S. K., Mackinnon,

J. M., Roberts, M. J., Buchanan, C. J., Kruczkiewicz,

P., Jokinen, C. C., Thomas, J. E., Nash, J. H. E.,

Gannon, V. P. J., Marshall, B., Pollari, F., and Clark,

C. G. (2012). Development and validation of a

comparative genomic ﬁngerprinting method for high-

resolution genotyping of Campylobacter jejuni. Jour-

nal of Clinical Microbiology, 50(3):788–797. PMID:

22170908.

van Belkum, A. (2003). High-throughput epidemiologic

typing in clinical microbiology. Clinical Microbiol-

ogy and Infection, 9(2):86–100.

Vogel, U., Szczepanowski, R., Claus, H., J

unemann, S.,

Prior, K., and Harmsen, D. (2012). Ion torrent per-

sonal genome machine sequencing for genomic typ-

ing of Neisseria Meningitidis for rapid determination

of multiple layers of typing information. Journal of

Clinical Microbiology, 50(6):1889–1894.

Ward, T. J., Usgaard, T., and Evans, P. (2010). A targeted

multilocus genotyping assay for lineage, serogroup,

and epidemic clone typing of Listeria monocytogenes.

Appl. Environ. Microbiol., 76(19):6680–6684.

MIST:AToolforRapidinsilicoGenerationofMolecularDatafromBacterialGenomeSequences

323