pdf to jpg c#: Create pdf security SDK control service wpf azure asp.net dnn jib-1840-part418

Automatic extraction of microorganisms and their habitats from

free text using text mining workﬂows

BalaKrishnaKolluru

1,2*

,SirintraNakjang

3,4

,Robert P.Hirt

,Anil Wipat

3,4

,Sophia

Ananiadou

1,2

NationalCentrefor TextMining,Universityof Manchester,131PrincessStreet,Manchester

M17DN,UK

SchoolofComputer Science, UniversityofManchester, KilburnBuilding,OxfordRoad,

Manchester,M139PL,UK

Institutefor CellandMolecular Biosciences,UniversityofNewcastle,Newcastle uponTyne,

NE24HH,UK

SchoolofComputingScience,UniversityofNewcastle,NewcastleuponTyne,NE17RU,UK

Summary

In this paper we illustrate the usage of text mining workﬂows to automatically extract

instances of microorganisms and their habitats from free text; these entries can then be

curatedandaddedtodifferentdatabases. Tothisend, we useaConditionalRandomField

(CRF) basedclassiﬁer,aspartof theworkﬂows,toextractthementionofmicroorganisms,

habitatsandtheinter-relationbetweenorganismsandtheirhabitats.

Results indicatea goodperformance forextractionof microorganisms andtherelationex-

tractionaspectsofthetask(withaprecisionofover80%),whilehabitatrecognitionisonly

moderate (a precision of about 65%). We also conjecture thatpdf-to-text conversioncan

bequitenoisyandthisimplicitlyaffectsanysentence-basedrelationextractionalgorithms.

1 Introduction

Microorganisms play a signiﬁcant role in symbiotic relationships with animal hosts ranging

from mutualism, commensalism to parasitism. To gain more insight into the mechanisms in-

volved in the host-microbeinteractions, it isessentialto beableto contrast genotypicfeatures

of microorganisms from various sources where microbes live, including both host-associated

(fromarangeofhosts-microbescontexts) and variousenvironmental niches[7]. To date, there

is no detailed datasourcefor this information regarding habitat or isolation source ofmicroor-

ganismswhosegenomesequencedataare available. GOLD[12] and NCBI[13] databasesare

someof the most popular publicresources whereinformation describing taxa can be obtained

in aformof ﬂatﬁles.

Due to thelarge numbers of taxafor which genomesequencedata are available, and their in-

creaseondailybasis, thereisanurgentneedto beabletodescribethehabitatorisolationsource

Towhomcorrespondenceshouldbeaddressed. Email: kollurub@cs.man.ac.uk

Journal of Integrative Bioinformatics, 8(2):184, 2011

http://journal.imbio.de

doi:10.2390/biecoll-jib-2011-184

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License (http://creativecommons.org/licenses/by-nc-nd/3.0/).

Create pdf security - C# PDF Digital Signature Library: add, remove, update PDF digital signatures in C#.net, ASP.NET, MVC, WPF

Help to Improve the Security of Your PDF File by Adding Digital Signatures

pdf security password; convert locked pdf to word

Create pdf security - VB.NET PDF Digital Signature Library: add, remove, update PDF digital signatures in vb.net, ASP.NET, MVC, WPF

Guide VB.NET Programmers to Improve the Security of Your PDF File by Adding Digital Signatures

change pdf document security; add security to pdf

for each taxon in an automated and consistent fashion. This issue was recently speciﬁcally

recognised bythe“Minimalinformation aboutaGenomeSequence”(MIGS)speciﬁcations[8]

and a fewpapershavediscussed this further orapplied someinitial approaches to address this

issue [9, 10]. In order to gather this scattered information, text-mining approaches are em-

ployed to extract metadata from the published literatures. Here, we present a biological use

case investigating the aspects ofmapping habitat to microorganismsviathe useoftext-mining

techniques.

Extracting microorganisms and their habitats are just a part of the information-processing

paradigmthatcouldinvolveseveral otherstages.

Weaddress two related issuesin thispaper:

1. Howeffectivearethestatisticalapproachesforextracting microorganismsandtheirhabi-

tats?

2. Develop workﬂows to combine text processing, named entity recognition and relation-

mining

2 Related work

Anumberofinvestigatorshaveworked on automaticentity extraction fortheBiology domain.

Hanisch et al. [1] have developed a system called ProMiner for identifying entities from sci-

entiﬁc literature. They have subsequently expanded this work for identifying proteins, genes

and diseases. Sasaki et al. developed a CRF based Named Entity Recogniser (NER), Nemine

that identiﬁesgenesandproteinsusing domain-speciﬁcdictionaries[4]. Usingfeaturessuchas

orthographic features, Part-of-Speech (POS) tags, dictionaries and contextual information an

F-scoreof 78.72% hasbeen achieved. Thesamemethod hasbeen adapted for the recognition

of metabolites (using the dictionary ChemSpider) achieving an F-score of 78.49% (precision

of83.02%)[15]. SimilarNER techniqueshavebeen deployed within theadvanced search ser-

vice, KLEIO [22], at the National Centre for Text Mining, UK. KLEIO uses NER boosted

by including term variation (acronym detection) and normalisation (spelling variants). Collier

et al. have developed a system called Biocaster [5], which employs ontologies and rules to

support text mining to track public health rumours. Biocaster reported an F-scoreof 76.97%.

Ananiadou et al. have employed named entity recognisers for type IV secretion systems and

shown theapplicabilityofstatisticalapproachesforthisdomain [21]. Theyreported an F-score

ofabout 90%fornamed entity recognition foridentiﬁcationof bacteria.

3 Data

Since this task represents novel challenges for application of text mining, there were no prior

standard annotated corporato serveastraining dataformachinelearning algorithmsorto pro-

vide a gold standard for evaluation. Furthermore, the types of relations and patterns of term

occurrence, in which we areinterested, arenot typically attested in theabstracts ofthepapers,

butonly appearaspartofthefulltextofthearticles. Therefore, wedevelopednewtraining and

evaluation corpusmaterialfortheseconceptsof interest, based on annotation of full papers. In

Journal of Integrative Bioinformatics, 8(2):184, 2011

http://journal.imbio.de

doi:10.2390/biecoll-jib-2011-184

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License (http://creativecommons.org/licenses/by-nc-nd/3.0/).

VB.NET PDF Password Library: add, remove, edit PDF file password

Set PDF security level. As String = Program.RootPath + "\\" 3.pdf" Dim outputFilePath As String = Program.RootPath + "\\" 3_pw_a.pdf" ' Create a password

decrypt pdf online; decrypt password protected pdf

C# PDF Password Library: add, remove, edit PDF file password in C#

Able to create a password protected PDF contains file permission limitation. your PDF document in C# project, XDoc.PDF provides some PDF security settings.

decrypt pdf password; decrypt a pdf file online

thesubsequentsections, wedescribethecriteriaused whilecreating thecorpus. Inbothclasses

of entity, i.e., microorganisms and habitats, we followed a similar paradigm, employing inde-

pendent manual annotation ofsomeamount of text and augmenting thatmaterial as necessary

through theuseof an “accelerated annotation”(Acela)interface[18].

Acela iteratively and interactively trains a machine learning classiﬁer to recognise a speciﬁc

concept or entity class, based on the current set of labelled examples, labels new example

sentences. Theseexamplesarethen giventothehuman annotatorforveriﬁcation orcorrection.

In previous work [18], thisapproach has been shown to achievefull annotation coveragewith

roughly50%oftheannotatoreffort, byfocusing on only thosesentences, whicharemostlikely

to includeitemsthat need to beannotated.

3.1 Organism-habitatcorpus

Forthiscorpus, two classesof entities wereannotated: microorganismsand habitats.

• Microorganisms: Scienticnamesofmicroorganisms, including bacteria, archaeaandmi-

crobialeukaryotes, areannotated

1. If they are speciﬁed at least to the genus level of precision. Species, strain, and

serovar entries are also tagged, if they are present. Typical examples of microor-

ganismsareCampylobacterspp., EscherichiacoliK12 andTrichomonasvaginalis.

2. Iftheyareinsentenceswhichcontain habitatorisolation sourceinformation forthe

organism.

• Habitats: Habitats or isolation sources of organisms, such as thosestored in the GOLD

database[12], aretagged

1. If they are context-related or can be referred to as a habitat or isolation source of

an organism. For example, if they refer to a host organism (human, cow), a body

partororgan of ahostorganism(lung, gut,lung abscess), refer to an environmental

habitat (minetailing, wastewater), or employ theadjectivalforms ofhabitatslisted

above(bovine, pulmonary, rumen).

2. Ifthey arenotassociated directly with diseasee.g. diarrhea, respiratory tractinfec-

tion.

3. Ifthey arein sentences, which contain theorganismassociated with thehabitat.

Afully annotated sentencewith organism and habitatinformation isshown here

Bacteroidessalyersaesp. nov. isolatedfromclinicalspecimensofhumanintestinal

origin.

An exampleof theexclusion criterion for organism annotation appearsin the sentence below,

where Campylobacter is not tagged because the sentence lacks correct context; there is no

habitatinformation.

The Campylobacter species were all isolated anaerobically and identiﬁed by se-

quencing analysisofthe 16S rRNA gene.

Journal of Integrative Bioinformatics, 8(2):184, 2011

http://journal.imbio.de

doi:10.2390/biecoll-jib-2011-184

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License (http://creativecommons.org/licenses/by-nc-nd/3.0/).

Online Split PDF file. Best free online split PDF tool.

into Multiple ones. You can receive the PDF files by simply clicking download and you are good to go!. Web Security. We have a privacy

secure pdf file; pdf security remover

Online Remove password from protected PDF file

If we need a password from you, it will not be read or stored. To hlep protect your PDF document in C# project, XDoc.PDF provides some PDF security settings.

can print pdf security; decrypt pdf password online

Table1: CorpusstatisticsforNewcastleOrganism-Habitatdata

Microorganisms

Habitats

Annotated sentences

1418

1609

Tokens(words)

57020

Tagged Tokens

40349

47393

TokensEntities

921

875

Tagged tokens(within theentities)

1951

1201

Estimated coverage

99.8%

99.4%

3.2 Corpus statistics

Here, the corpuswasseeded with aset ofannotated excerptsfrom22 fullpapersthat speciﬁed

organismsandhabitats, drawn fromabibliography provided byadomainexpert. An additional

set of10 fulltext documentsfrom thesameexpertwasemployed in theaccelerated annotation

process. Annotator instances for the two classes, microorganism and habitat, were created.

An expert annotator then interactively applied the annotation interfaceto label instances until

an estimated coverage over 99% was achieved. Estimated coverage is a ratio of number of

manually annotated entities to the total number of entities as expected by the CRF classiﬁer.

See[18] fordetails. Detailed statistics fortheresulting corpusarepresented in Table1.

4 Our approach

In thispaper,weaddresstheissueofautomaticallyextractingdifferentorganisms,theirhabitats

and the inter-relation from free literature. We employ workﬂows using text-mining tools for

identiﬁcation oforganismsand habitats. Wehaveused U-Compare[3]to design aworkﬂowto

build a named entity recogniser to accept plain text or a pdf. This workﬂow can be run both

in the U-Compare environment as a stand-alone or as part of Taverna workﬂow management

system[2]. Figure1 showstheschematicsview ofourworkﬂow.

Figure1:Overviewofourapproach.

Thepdf-to-textconvertercomponentoftheworkﬂowwasbuiltontheapplication programming

Journal of Integrative Bioinformatics, 8(2):184, 2011

http://journal.imbio.de

doi:10.2390/biecoll-jib-2011-184

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License (http://creativecommons.org/licenses/by-nc-nd/3.0/).

C# HTML5 Viewer: Deployment on AzureCloudService

All. Create a New AzureCloudService Project in RasterEdge.XDoc.PDF.HTML5Editor.dll. validateIntegratedModeConfiguration="false"/> <security> <requestFiltering

convert secure pdf to word; pdf security

C# HTML5 Viewer: Deployment on ASP.NET MVC

Create a New ASP.NET MVC3 RasterEdge.XDoc.PDF.HTML5Editor.dll. validateIntegratedM odeConfiguration="false"/> <security> <requestFiltering allowDoubleEscaping

change security on pdf; add security to pdf file

interface(API) provided by the Apachegroup [19]. The component reads individual ﬁle and

convertsinto to raw text.

The workﬂow implementing the named entity recogniser employs conditional random ﬁelds,

CRFs [6] using a combination of dictionary (NCBI/habitat list) features, lexical features, or-

thographic features and contextual features. CRFsarea type of discriminative and undirected

probabilitiesgraphicalmodelsoftenused fortagging sequentialdataand innamedentityrecog-

nition in naturallanguageprocessingand biologydomains[16, 17]. In ourimplementation, we

usetheMallet [23]implementation of CRFswith lexicaland orthographicfeaturesto train the

CRF model. We have also employed two dictionaries that were tailored to the task from a

combination of established and curated domain ontologies and term lists provided by domain

experts.

4.1 Resources

• Microorganism Resources Two large-scale resources for scientic names for microor-

ganismswereused:

1. Microorganisms’ scientiﬁc names from NCBI taxonomy including bacteria, ar-

chaea, microbial eukaryote (http://ncbi.nlm.nih.gov/taxonomy)

2. ListofProkaryoticnamesandmicrobialeukaryoteswithStandinginNomenclature

(LPSN)(http://www.bacterio.net)

Thebacterialnamesfromtheseresourceswereconvertedto asetofstandardised formsto

covertypicalvariability fortheseterms including thecommon abbreviation of thegenus

term, removalofspecies, strain, orserovarcomponents,removaloftagssuch as‘subsp.’,

‘str.’, ‘strain’, etc., and generation of plural forms for genus terms. The resulting term

listcomprises52715 entriesfor12256 distinct organisms.

• HabitatResourcesA key source ofhabitat termsisthe GOLD [12] database, which in-

cludes135habitat types. After somesimplenormalisation to enhancematchingbetween

the terms and running text, the speciﬁc dictionary was further enhanced with names of

animals(169306entries, 166244 head terms)and body partsandorgans(120668 entries,

56578 head terms) extracted from theUMLS Metathesaurus[14].

4.2 Features forthe CRF model

These experiments incorporate three main sets of base features, inspired by previousresearch

in biomedicalNER [4].

• Lexical features are current word, the root form of the current word, and the part-of-

speech tag ofthecurrent word, computed by the Genia tagger[24].

• Orthographicfeaturesaremadeupofsubstringandwordformfeatures. In theword form

features, all uppercase letters are converted to ‘A’, lowercase to ‘a’, and all numbers to

‘0’. The ﬁrst two & four characters and the last two & four characters of the original

word and theword form arechosen as features.

Journal of Integrative Bioinformatics, 8(2):184, 2011

http://journal.imbio.de

doi:10.2390/biecoll-jib-2011-184

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License (http://creativecommons.org/licenses/by-nc-nd/3.0/).

VB.NET PDF Library SDK to view, edit, convert, process PDF file

allows users to perform PDF document security settings in added to a specific location on PDF file page. In addition, you can easily create, modify, and delete

cannot print pdf security; secure pdf

VB.NET PDF Page Extract Library: copy, paste, cut PDF pages in vb.

Security PDF component download. You can easily get pages from a PDF file, and then use these pages to create and output a new PDF file.

decrypt pdf; convert locked pdf to word doc

• Dictionary featuresarebinary features to indicatethepresenceof theword in thedictio-

nary and theposition ofthewordwithin any dictionary entries.

For each of the base features, corresponding features for words within a context window are

added to the representation. The window ranges from 1-3 words preceding and following the

currentword.

4.3 Hybrid Dictionary-Machine Learning Based Approach

For entity recognition, we used the CRF based approach. The current approach employs a

sequenceclassiﬁer, trainedon ahand-annotatedcorpus,which wasconvertedtoastandardBIO

(Begin of a sequence, Inside a sequence, Outside a sequence) format. This corpus consisted

of32 full papers from various journalsand wasspeciﬁcally annotated for microorganisms and

habitats. TheCRFs were employed with alinearchain model.

4.4 Machine Learning Approach for Relation mining workﬂows

The main focus of this experiment was to elicit the sentences containing a microorganism-

habitatrelation fromfree-text, typically in apdf-format.

Theworkﬂowhad threeprincipal components, asshown in Figure1:

1. PDF-to-text convertor: this component was based on Apache’s pdfbox API. It converts

each pdfinto astream oftext intoa JavaString object.

2. Named entity recogniser, as described in the earlier section: input from this component

was used to populatethefeature-space forCRF component.

3. CRF-componentforbuilding abespokerelation mining apparatus

Relation miningworkﬂow wasbased onaCRF modeltrained onco-occurrenceofmicroorgan-

isms and their habitats, thus modelling the relation between a microorganism and its habitat.

Just as forthenamed entity recognition experiments described in earlier sections, Mallet [23]

implementation was used for these experiments with a feature-space designed for sentences

extraction.

4.5 Features forRelation mining CRF component

Theseexperimentsincorporatetwo main setsofbasefeatures

• Entity occurrencein asentencei.e. ifasentencehad any organismorhabitat

• Contextual information (theword preceding theentity and theword following it)around

theoccurred entities

Journal of Integrative Bioinformatics, 8(2):184, 2011

http://journal.imbio.de

doi:10.2390/biecoll-jib-2011-184

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License (http://creativecommons.org/licenses/by-nc-nd/3.0/).

Table2:PerformanceofCRFanddictionary-basedapproaches

Microorganisms

Habitats

precision

)

recall

)

F-score

)

Dictionary

CRF

5 Results & Discussion

5.1 Dictionary-based & CRF approaches

First, we consider the contrast between recognition accuracy for the systems employing the

CRF with allfeatures trained on theannotated corpora with therecognition accuracy obtained

using a simple, longest match strategy using dictionary resources. The results are shown in

Table2.

We can observe that the results range from F-score of 56% to 63% for dictionary-based ap-

proaches and from 57%to 81%for machinelearning based methods, using all features. These

ﬁgures show some interesting contrasts. For microorganism tagging, the CRF approach out-

performs the dictionary. While recall remains good for the dictionary-based approach, the

precision ison 54%. Inotherwords, the dictionary-based approach had ahigh numberof false

positives. Theprecision improved to 84% in theCRF setup; even though therecallisnotvery

differentfromthedictionary-based approach.

Furthermore,thehabitatannotationtaskisshownto beparticularly challenging. Severalfactors

such as a very broad class deﬁnition spanning animals, anatomy, environment and adjectival

forms thereof, in conjunction with the restriction to organismal contexts and alack of explicit

word-morphologies restrict theCRF-based approach.

5.2 Hybrid CRF

As can be seen from Table 3, the hybrid classiﬁer has achieved an F-score of about 80% and

60%for OrganismsandHabitats respectively on a9-fold cross validation.

Asa general observation, microorganisms had a distinct attributes: word shape features, such

asuppercaseletters, asubtlepattern innamessuch as“ccus”or “cci”etc. Thecurrentfeatures

that wehave used, could model themicroorganismswell and an F-Score of81 isan indicative

ofthissuccess. Typicalfalsenegativesincludehyphenatedwordssuch as“B. taylorii-like”and

partial names such as “M. succiniciproducens” of the complete name “M. succiniciproducens

MBEL55E(KCTC0769BP)”.

Habitats, on the other hand, seemed to lack any such explicit attributes making them more

difﬁcult to model. There were quite a few falsenegatives such as water, skin, abcessetc. We

conjecture that perhapsmoretraining data(weused only 850 training instances ofhabitatsfor

theexperimentsdiscussed here)could certainly helpalleviatesuch falsenegatives. Wearealso

currentlyexploring wideningour feature-spaceby including deeper semanticinformation such

asparse-treeinformation to enhancetheperformanceofhabitat recognition.

Journal of Integrative Bioinformatics, 8(2):184, 2011

http://journal.imbio.de

doi:10.2390/biecoll-jib-2011-184

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License (http://creativecommons.org/licenses/by-nc-nd/3.0/).

Table3:Performanceofourapproach

Classofentities

Precision (%)

Recall(%)

F-score(%)

Microorganism

Habitats

Table 4: Performanceofrelationextractionapproach

Precision (%)

Recall (%)

F-score(%)

Relations

6 Relation mining

Therelation mining classiﬁerachieved aprecision ofabout 85% and arecall ofless than 50%

(Table 4). We conjecture that slightly below parperformanceof habitats waspartly accounted

for by the contextual features as a result of which the precision was not proportional to the

individual precisions of microorganisms and habitats. One of the important factors for the

number of false negatives is the noise generated from pdf-text conversion. Typical examples

of“corrupted”sentencesincludethosewheresentenceboundariescould notbeidentiﬁed auto-

matically and thereforehad morenoisy features. This could havehad an impact on the perfor-

mance. A typicalexampleofcorrupted sentenceisajournaltitleconcatenatingwith thetitleof

thepaperisshown here(taken from [25])

“Biometrika 40:237–264 Hallberg KB, Johnson DB (2003) Novel acidophiles iso-

lated frommoderatelyacidicminedrainagewaters.”

And another example where the caption of a ﬁgure and some text from the ﬁgure itself are

combined with thetextfrom[26]isshown here:

“These mechanisms may have evolved in bacterial pathogens to increase the fre-

quency of phenotypic variation in genes involved in 1 100,000 200,000 300,000

1,600,00 Figure2 Circular representation oftheH. pylori26695 chromosome.”

Clearly, any paradigm that extracts sentences will be affected by such erroneous conversion.

Thereareseveralsoftwareswhichconvertapdf-documenttotextsuchasApache’spdfbox[19],

Utopia[20]and Unixcommand “pdftotext”. Butallofthemhavethesameproblemsin dealing

with tablesandﬁguresinpdfs,whichresultin suchnoise. Wearecurrentlyworkingeliminating

someoftheerroneous sentencesusing statistical methodssuch as languagemodelling. Weare

also working on using parsers to eliminate some of HTML-related noisesuch as captions and

headers etc.

Anotherfactorthatwecould identifyforthelower recall wasthebelow parperformanceofthe

habitatrecognition by theautomaticnamed entity recognisers.

6.1 A toy example

Consider asmallexcerpt aboutAcinetobacter species [27]

Journal of Integrative Bioinformatics, 8(2):184, 2011

http://journal.imbio.de

doi:10.2390/biecoll-jib-2011-184

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License (http://creativecommons.org/licenses/by-nc-nd/3.0/).

“Acinetobacterspparewidelydistributed in natureand soil. Mostinfectionsoccur

in immunocompromised individuals, and thestrain Acinebacter baumannii is the

second most commonly isolated nonfermenting bacteria in human specimens. It

can surviveon the human skin ordry surfacesforweeks.”

Figure2:TheresultofourworkﬂowonextractionofmicroorganismsandhabitatsinU-Compare

Figure2 showstheresultofour text-mining workﬂowon this excerpt. All themicroorganisms

areunderlined in red, thehabitats areunderlined in green and therelation-indicating sentences

areunderlined in yellow. Theworkﬂowunderlinessentencesin yellow ifand only ifitcontains

both microorganisms and their habitats, thus highlighting host-associated habitats from non-

host-associated ones.

7 Conclusion

Ascanbeinferred fromtheresults, theworkﬂowsthatwehavedeveloped achieveareasonable

accuracy thatmakesthemlikely candidatestobeported forotherdatabasesand similarapplica-

tions. Weareplanning to implementthisworkﬂowto curateamicroorganism-habitatdatabase.

As the workﬂows are inter-operable as sub-workﬂows in Taverna [2], we hope the scientiﬁc

community at large, and biologists in particular, can make use of them for their respective

research projects.

Acknowledgements

Thisresearch hasbeen supportedby theBiotechnology &BiologicalSciencesResearch Coun-

cil (ONDEX project, BB/F006039/1). The National Centre for Text Mining is funded by the

JointInformation SystemsCommittee. SirintraNakjang wassupportedby theFaculty ofMed-

ical Sciences and theSchool of Computing Science at Newcastle University and an Overseas

Research StudentsScheme.

Journal of Integrative Bioinformatics, 8(2):184, 2011

http://journal.imbio.de

doi:10.2390/biecoll-jib-2011-184

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License (http://creativecommons.org/licenses/by-nc-nd/3.0/).

References

[1] D. Hanisch, K. Fundel, H.-T. Mevissen, R. Zimmer and J. Fluck. ProMiner: Organism

speciﬁc protein name detection using approximate string matching. EMBO Workshop,

Granada, Spain, March 28–31, 2004.

[2] Taverna, http://www.taverna.org.uk, last accessed on 08 July 2011.

[3] Y. Kano, W. A. Baumgartner, L. McCrohon, S. Ananiadou, K. B. Cohen, L. Hunder and

J. Tsujii. U-Compare: share and comparetext mining toolswith UIMA. Bioinformatics,

25(15):1997–1998, 2009.

[4] Y. Sasaki, Y. Tsurouka, J. McNaught and S. Ananiadou. How to make the most of NE

dictionariesin statisticalNER. BMCBioinformatics, 9(Suppl 11):S5, 2008.

[5] N. Collier, S. Doan, A. Kawazoe, R. M. Goodwin, M. Conway, Y. Tateno, Q-H. Ngo, D.

Dien, A. Kawtrakul, K. Takeuchi, etal. BioCaster: detecting publichealth rumorswitha

Web-based textmining system.Bioinformatics, 24(24):2940–2941, 2008.

[6] J. Lafferty, A. McCallumand F.Pereira.Conditional randomﬁelds: Probabilisticmodels

for segmenting and labeling sequencedata. In Proceedingsof 18th InternationalConfer-

enceon MachineLearning, 2001.

[7] O.O’Sullivan, J. O’Callaghan, A.Sangrador-Vegas, O. Auliffe, L. Slattery, P. Kaleta, M.

Callanan, G. F. Fitzgerald, R.P. Ross, and T. Beresford. Comparativegenomics of lactic

acid bacteriarevealsa niche-speciﬁcgeneset. BMCMicrobiology, 9(1):50, 2009.

[8] D.Field,G.Garrity,T.Gray, N.Morrison,J. Selengut,P.Sterk, T. Tatusova,N.Thomson,

M. J. Allen, S. V. Angiuoli, et al. The minimum information about a genome sequence

(MIGS)speciﬁcation. NatureBiotechnology, 26(5):541–547, 2008.

[9] L. Hirschman, C. Clark, K. B. Cohen, S. Mardis, J. Luciano, R. Kottman, J. Cole, V.

Markowitz, N. Kyrpides, N. Morrison, et al. Habitat-Lite: a GSC case study based on

freetexttermsforenvironmental metadata. OMICS, 12(2):129–136, 2008.

[10] C. von Mering, P. Hugenholtz, J. Raes., S. G. Tringe, T. Doerks, L. J. Jensen, N. Ward,

and P. Bork. Quantitative phylogenetic assessment of microbial communities in diverse

environments. Science, 315(5815):1126–1130, 2007.

[11] J. H. Lee, V. N. Karamychev, S. A. Kozyavkin, D. Mills, A. R. Pavlov, N. V. Pavlova,

N. N. Polouchine, P. M. Richardson, V. V. Shakhova, A. I. Slesarev, et al. Comparative

genomicanalysisof thegut bacterium longumrevealslocisusceptibleto deletion during

pureculturegrowth. BMC Genomics, 9(1):247, 2008.

[12] K.Liolois, I. A. Chen,K.Mavromatis, N. Tavernarakis, P.Hugenholtz,V.M.Markowitz,

and N. C. Kyrpides. The Genomes On Line Database (GOLD) in 2009: status of ge-

nomicand metagenomicprojectsand their associated metadata. Nucleic AcidsResearch,

38(Databaseissue):D346–D354, 2010.

Journal of Integrative Bioinformatics, 8(2):184, 2011

http://journal.imbio.de

doi:10.2390/biecoll-jib-2011-184

This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License (http://creativecommons.org/licenses/by-nc-nd/3.0/).

Documents you may be interested

pdf to jpg c# : Create pdf security SDK control service wpf azure asp.net dnn jib-1840-part418