Update on XplorMed: A web server for exploring scientific literature

Abstract

As scientific literature databases like MEDLINE increase in size, so does the time required to search them. Scientists must
frequently inspect long lists of references manually, often just reading the titles. XplorMed is a web tool that aids MEDLINE
searching by summarizing the subjects contained in the results, thus allowing users to focus on subjects of interest. Here
we describe new features added to XplorMed during the last 2 years (http://www.bork.embl-heidelberg.de/xplormed/).

Full text

Update on XplorMed: a web server for exploring
scientific literature
Carolina Perez-Iratxeta
1,2,
*, Antonio J. Pe
´rez
3
, Peer Bork
1,2
and Miguel A. Andrade
1,2
1
European Molecular Biology Laboratory, Meyerhofstr. 1, 69117 Heidelberg, Germany,
2
Max Delbru¨ck
Center for Molecular Medicine, Robert Ro
¨sler-Str. 10, 13125 Berlin-Buch, Germany and
3
University of Malaga,
Facultad de CC. Campus de Teatinos, 29071 Malaga, Spain
Received February 14, 2003; Revised and Accepted March 21, 2003
ABSTRACT
As scientific literature databases like MEDLINE
increase in size, so does the time required to search
them. Scientists must frequently inspect long lists of
references manually, often just reading the titles.
XplorMed is a web tool that aids MEDLINE searching
by summarizing the subjects contained in the
results, thus allowing users to focus on subjects of
interest. Here we describe new features added to
XplorMed during the last 2 years (http://www.bork.
embl-heidelberg.de/xplormed/).
BACKGROUND AND GOALS
A scientist searching the scientific literature (for example,
MEDLINE with Entrez at the NCBI’s PubMed server, http://
www.ncbi.nlm.nih.gov/entrez/) may initially retrieve an
unmanageable number of references, typically hundreds.
Even for very specific subjects, it is not always clear how to
narrow the search to focus on the most relevant matches. For
example, imagine you are a researcher interested in the
possible role of the interaction between heparin and proteins in
Alzheimer’s disease, who queries the PubMed server with the
terms ‘Alzheimer and heparin’. This search returns presently
>100 references to literature, of which only some mention
proteins. Finding these currently requires manual examination
of the abstracts which can be time-consuming. In such
instances XplorMed can be useful (1).
XplorMed is a web tool that summarizes MEDLINE search
results according to subjects and allows you to navigate
through abstracts in an interactive fashion. Here we give details
as to the use of XplorMed. A detailed tutorial is also avail-
able online (http://www.bork.embl-heidelberg.de/xplormed/
example/).
INPUT TO XplorMed
There are two ways to provide input to XplorMed. You can
type a PubMed query directly into our server or you can supply
a file containing a set of abstracts. XplorMed can handle
several abstract formats: MEDLINE (default), EndNote, XML
and XplorMed (see page 1 of the tutorial for details).
A third way to query XplorMed is to start from literature
linked to a particular entry from one of the MEDLINE, OMIM
(2), SMART (3) or SWISS-PROT/SpTrEMBL (4) databases.
Here you simply need to provide the identifier of the entry of
your interest and the corresponding database name. The initial
set of abstracts of each XplorMed session is kept in the server
for a week, enabling you to recover your session. We
recommend you start with sets of 30 references, though the
current maximum is 500 abstracts.
OVERVIEW OF AN ANALYSIS
The first step involves a coarse overlapping clustering of the
abstracts. References are classified into eight classes depend-
ing on their subject. Classes correspond to MeSH main
categories, such as ‘Anatomy’, ‘Organisms’, ‘Chemical and
Drugs’, ‘Biological Sciences’, etc. (see http://www.nlm.nih.
gov/mesh/meshhome.html). You can impose an initial filtering
to restrict the search to categories of interest and it is also
possible to filter the search results by publication date (see
page 2 of the tutorial).
The next web page displays keywords in the selected
abstracts. The method for computing keywords and relations
between them can be found in literature (5). The list of
extracted keywords provides a summary of the subjects within
the query results and these are listed in order of relevance
(more important concepts are listed first). Considering the
above example of heparin and Alzheimer, XplorMed gives
expected terms—‘protein’, ‘heparin’, ‘alzheimer’ and
‘disease’—in addition to others that may be new to you,
for example, ‘tau’ and ‘app’.
At this stage, you can choose whether to go directly to the
next step or to start a deeper analysis of the displayed subjects.
The latter involves a context analysis of the subjects
represented by the keywords and it is outlined briefly below
(see Context Analysis of the Subjects). Alternatively, if you
choose to go further, several groups or chains of closely related
keywords are then presented to you.
You can modify the number of chains and their length by
means of two parameters: alpha and score (see page 3 of
*To whom correspondence should be addressed. Tel: +49 6221 387 456; Fax: +49 6221 387 517; Email: cperez@embl-heidelberg.de
3866–3868 Nucleic Acids Research, 2003, Vol. 31, No. 13
DOI: 10.1093/nar/gkg538
Published by Oxford University Press 2003

Figure 1. (A) XplorMed’s home page. (B) Words related to ‘app’. (C) Sentences containing the words ‘app’ and ‘outgrowth’.
Nucleic Acids Research, 2003, Vol. 31, No. 13 3867

tutorial for details). Each chain is preceded by a number that
indicates how many abstracts contain both words. By selecting
one or more of these chains, you perform a sub-query of the
original set. For example, suppose you are interested in protein
domains that could bind heparin. Accordingly, you would
inspect the pair {protein, domain}, which appears in 13
references. You can select an alternative or additional word
chain if you do not find what you wanted among the proposals
of the system.
The next web page provides an ordered list of abstracts;
those likely to be most interesting according to your selection
are highlighted on top (in our example, the papers dealing with
the heparin binding domain). If you checked in the previous
page the boxes for cross-linking to the corresponding
databases, several hyperlinked symbols will label some
abstracts (see Cross Linking to Molecular Biology Databases).
The filtered subset of papers can now be used as a new
XplorMed starting point at the computation-of-keywords step
(see above). Alternatively, you can expand this subset with
new papers among their MEDLINE neighbors (see Expanding
the Query through Related Bibliography). New keywords
focusing more closely on your subject of interest will appear at
this stage. The procedure can be performed repetitively and the
recovery of the set of abstracts is possible at any stage.
CONTEXT ANALYSIS OF THE SUBJECTS
When the list of keywords is presented, you can explore both
their meanings and relationships. By clicking on a word you
can see all the sentences in the abstracts that contain that word
and each sentence is linked to its MEDLINE abstract. In this
way you can learn why a particular word is mentioned across
the abstracts. Moreover, you can also discover interesting
information by examining the words strongly related to a
particular word (for example, ‘app’, see Fig. 1B). By clicking
on the [R] next to each word, a window displaying closely
related words (such as ‘outgrowth’or ‘zinc’) will be shown.
Clicking on the [X] near any related word (like ‘outgrowth’)
shows the sentences containing either of the words (‘app’or
‘outgrowth’) in abstracts containing both words [for example,
‘The results indicate that the binding of APP to HSPG in the
ECM may stimulate the effects of APP on neurite outgrowth.’
(6)]. Words and sentences are highlighted in different colors
for an easy identification (see page 3 of the tutorial for details).
Clicking the button ‘Explore the context of any word’allows
you to do this kind of analysis in a more flexible way by typing
other keywords of interest.
CROSS-LINKING TO MOLECULAR BIOLOGY
DATABASES
As was mentioned above, the list of selected abstracts can be
optionally hyperlinked to objects in several databases,
currently MEDLINE, OMIM, SMART, SWISS-PROT and
SpTrEMBL. The diverse symbols indicate the database and in
the case of SWISS-PROT, the subject of the article, such as
‘describes protein function’,‘reports a 3D structure’, etc. Note
that the hyperlink to PubMed is always supplied, allowing you
to check the content of the abstract. An additional symbol
denotes review articles.
EXPANDING THE QUERY THROUGH
RELATED BIBLIOGRAPHY
As mentioned above, once you have selected a subset of
abstracts, it is possible to re-enter the analysis with the filtered
set at the computation-of-keywords step. You can also expand
this set of abstracts by retrieving neighbors from MEDLINE.
Neighbors are those references that deal with the same
(or similar) subject (7). To opt for this expansion you have
to check the box at the bottom of the list of references. You
can also change the number of neighbors included.
CONCLUSION
We have summarized how you can use the web tool XplorMed
to deal more efficiently with MEDLINE literature. Because
our server is being continually developed for the inclusion of
new features, any suggestion from users is warmly welcomed
and will be acknowledged.
ACKNOWLEDGEMENTS
We are grateful to the members of our group for their
suggestions and to Robert B. Russell and to Sea
´nI.
O’Donoghue for comments to our manuscript. XplorMed uses
in one step, TreeTagger, a part of speech tagger. We are
grateful to Helmut Schmid (IMS, Stuttgart, Germany) for
developing TreeTagger and making it publicly available.
REFERENCES
1. Perez-Iratxeta,C., Bork,P. and Andrade,M.A. (2001). Xplormed: a tool for
exploring MEDLINE abstracts. Trends Biochem. Sci.,26, 573–575.
2. Online Mendelian Inheritance in Man, OMIM (TM). McKusick-Nathans
Institute for Genetic Medicine, Johns Hopkins University (Baltimore, MD)
and National Center for Biotechnology Information, National Library of
Medicine (Bethesda, MD), 2000.
3. Letunic,I., Goodstadt,L., Dickens,N.J., Doerks,T., Schultz,J., Mott,R.,
Ciccarelli,F., Copley,R.R., Ponting,C.P. and Bork,P. (2002) Recent
improvements to the SMART domain-based sequence annotation
resource. Nucleic Acids Res.,30, 242–244.
4. Boeckmann,B., Bairoch,A., Apweiler,R., Blatter,M.C., Estreicher,A.,
Gasteiger,E., Martin,M.J., Michoud,K., O’Donovan,C., Phan,I. et al.
(2003) The SWISS-PROT protein knowledgebase and its supplement
TrEMBL in 2003. Nucleic Acids Res.,31, 365–370.
5. Perez-Iratxeta,C., Bork,P. and Andrade,M.A. (2002). Computing fuzzy
associations for the analysis of biological literature. Biotechniques,32,
1380–1385.
6. Small,D.H., Nurcombe,V., Reed,G., Clarris,H., Moir,R., Beyreuther,K.
and Masters,C.L. (1994). A heparin-binding domain in the amyloid
protein precursor of Alzheimer’s disease is involved in the regulation of
neurite outgrowth. J. Neurosci.,14, 2117–2127.
7. Wilbur,W.J. and Yang,Y. (1996). An analysis of statistical term strength
and its use in the indexing and retrieval of molecular biology texts.
Comput. Biol. Med.,26, 209–222.
3868 Nucleic Acids Research, 2003, Vol. 31, No. 13