Genomics 108 (2016) 151–157

Contents lists available at ScienceDirect

Genomics

journal homepage: www.elsevier.com/locate/ygeno

A PubMed-wide study of endometriosis

Ji-Long Liu ⁎, Miao Zhao
College of Veterinary Medicine, South China Agricultural University, Guangzhou 510642, China

a r t i c l e i n f o a b s t r a c t

Article history: Endometriosis affects 5–10% of women in reproductive age, leading to dysmenorrhea, pelvic pain and infertility;
Received 14 June 2016 however, our understanding on the pathogenesis of this disease remains incomplete. In the present study, we
Received in revised form 30 September 2016 performed a systematic analysis of endometriosis-related genes using text mining. Taking text mining results
Accepted 12 October 2016
as input, we subsequently generated a filtered gene set by computing the likelihood of finding more than expect-
Available online 13 October 2016
ed occurrences for every gene across the disease-centered subset of the PubMed database. Characterization of
Keywords:
this filtered gene set by gene ontology, pathway and network analysis provides clues to the multiple mechanisms
Endometriosis hypothesized to be responsible for the establishment of ectopic endometrial tissues, including the migration, im-
Pathogenesis plantation, survival and proliferation of ectopic endometrial cells. Finally, using this gene set as “seed”, we
Text mining scanned human genome to predict novel candidate genes based on gene annotations from multiple databases.
Gene prioritization Our study provides in-depth insights into the pathogenesis of endometriosis.
© 2016 Elsevier Inc. All rights reserved.

1. Introduction
Endometriosis is a common gynecological disorder, defined as the
presence of endometrial tissue outside the uterine cavity, primarily on
the pelvic peritoneum and ovaries. It affects 5–10% of women in repro-
ductive age and clinical manifestations are dysmenorrhea, pelvic pain
and infertility [1]. The pathogenesis of endometriosis is not fully under-
stood. Several hypotheses have been suggested, such as retrograde
menstrual reflux [2], immune system defects [3,4], and ectopic presence
of endometrial stem cells [5]. Notably, menstrual shedding is likely re-
quired for the development of endometriosis, as endometriosis occurs
spontaneously only in humans and some non-human primates [6].
Studies in twins have confirmed the involvement of genetic factors
in the etiology of endometriosis [7,8]. To elucidate causal genetic vari-
ants underlying endometriosis, many genes with an inferred biological
relevance to endometriosis have been chosen and genetic variants in
these genes have been tested for association with the disease [9]. As un-
biased approaches, several genome-wide association studies (GWAS)
have also been performed and a subsequent meta-analysis has revealed
6 loci consistently associated with susceptibility to endometriosis [10].
Additionally, extensive investigations using conventional methods
(e.g. Western blot) and high throughput methods (e.g. microarray)
have been performed to characterize gene expression differences be-
tween the eutopic and ectopic endometrium in order to better under-
stand and define the molecular basis of the disease [11]. Despite a
⁎ Corresponding author at: College of Veterinary Medicine, South China Agricultural
University, No. 483 Wushan Road, Tianhe District, Guangzhou 510642, China.
E-mail address: jilongliu@scau.edu.cn (J.-L. Liu).
large number of genes are identified, our understanding on the patho-
genesis of this disease remains incomplete.
A wealth of information is hidden within the large number of molec-
ular measurements from published experimental results. In the present
study, we performed a text mining analysis of endometriosis-associated
genes. A filtered gene set was subsequently generated by using
hypergeometric test. Analysis of this filtered gene set provides in-
depth insights into molecular mechanisms underlying the pathogenesis
of endometriosis.
2. Methods
2.1. Text mining
The PubMed database was used as a source of literature for text min-
ing. We conducted a search with the following combinations of query
key words: “endometriotic” OR “endometriosis” OR “endometrioma”
OR “endometriomas”. The relevant articles were retrieved in XML for-
mat, which makes information extraction more precise due to presence
of content enclosed within XML tag pairs. For each article, titles and ab-
stract texts were fetched and transformed into the PubTator format [12,
13] using in-house PERL scripts.
Text mining was performed by using the GNormPlus pipeline [14].
GNormPlus includes two modules: gene mention recognition and
gene name normalization, respectively. Gene mentions are detected
by using conditional random fields (CRF) model provided by CRF++ li-
brary. Gene mention recognition is integrated with species name recog-
nition by using SR4GN [15]. GenNorm [16], in combination with the
composite mention simplification tool SimConcept [17,18] and the

http://dx.doi.org/10.1016/j.ygeno.2016.10.003
0888-7543/© 2016 Elsevier Inc. All rights reserved.
152 J.-L. Liu, M. Zhao / Genomics 108 (2016) 151–157

Fig. 1. Identification of endometriosis-related genes by text mining. (A) Overview of the text mining pipeline. (B) The cumulative number of publications related to endometriosis from
1980 to 2016. (C) Distribution of the number of publications per gene.

abbreviation resolution tool Ab3P [19], is used for gene name normali- computed text mining results for the whole PubMed database at
zation. Only human genes are considered. By integrating these ad- the PubTator website [13]. We assumed that the total number of
vanced tools, GNormPlus pipeline compares favorably to other state- publications in the PubMed database is N and the number of publica-
of-the-art methods by achieving 87.1% precision and 86.4% recall [14]. tions on this particular gene is m. Then, the hypergeometric test has
Despite the performance was high, errors remained in the out- the following form:
put of GNormPlus. For example, IVF, which is short for in vitro fer-
tilization, was recognized as a gene mention and linked to SCN5A
X
k−1
gene (sodium channel voltage-gated type V alpha subunit) by p ¼ 1− pðijn; m; NÞ
mistake. To ensure accuracy, each entry from GNormPlus output i¼0
was checked manually. We identified 4381 false positive errors
out of a total of 38,490 entries. Thus, the observed precision was
where p(i | n,m,N) is the probability of observing exactly i publica-
4381/38,490 = 88.62%. Finally, all the curated entries were sum-
tions in the endometriosis-centered subset of the PubMed database:
marized and a full gene list associated with endometriosis was
compiled.
n!ðN−nÞ!m!ðN−mÞ!
pðijn; m; NÞ ¼
ðn−iÞ!i!ðn−mÞ!ðN−n−m þ iÞ!N!
2.2. Hypergeometric test

For a certain gene, the frequency of its occurrence in the text
mining results is denoted as k. Let the total number of publication
in the text mining results be n. Endometriosis-centered publications
consist only a small part of the PubMed database. In order to esti-
mate background distribution, we downloaded GNormPlus pre-
Finally, we calculated the significance threshold for multiple testing
using Bonferroni multiple test correction. In this way, the genome-wise
error rate is 1 − (1 − αi)n ≈ αin where αi is the individual test rejection
level and n is the number of genes under consideration. The algorithm
was implemented in PERL code.

Fig. 2. Filtering of text miming results by hypergeometric test. (A) A flowchart of the filtering process. (B) A Manhattan plot showing filtered results. The genomic location is ticked on the
x-axis and the negative decadic logarithm of the raw p-value is plotted on the y-axis. The dashed horizontal line is the threshold for genome-wide significance at p = 0.01 after Bonferroni
multiple test correction.
 J.-L. Liu, M. Zhao / Genomics 108 (2016) 151–157 153

2.3. Gene ontology, pathway and network analysis 2.4. Novel candidate gene prioritization

Gene ontology (GO) analysis was performed by using BiNGO ver-
sion 2.3 with GOslim database [20]. To test for enrichment, a
hypergeometric test was conducted followed by customized
Bonferroni multiple test correction. The adjusted p-value of 0.01
was used as significance threshold to identify enriched GO terms.
The R package wordcloud was used to generate word cloud for sig-
nificantly enriched GO terms. The font sizes in the word cloud
were proportional to − log10 of adjusted p-value for each enriched
GO terms. The DAVID tool [21] was employed for pathway enrich-
ment analysis. The same significance cutoff as GO analysis was
adopted. The STRING database version 10.0 [22] was used to create
gene network. The minimum combined score was set to 0.9. The
Cytoscape software was applied for visualization and analysis of
the gene network. The threshold of the degree value for hub genes
was defined as the mean plus two standard deviations.
We utilized the Endeavour software [23] for gene prioritization
throughout the human genome. Endeavour integrates multiple hetero-
geneous data sources into a global ranking using order statistics. The
order statistics based p-value, which represents the probability that a
candidate gene would obtain its rank by chance, is highly dependent
on the number of candidate genes considered. Because there is no rea-
sonable threshold for p-values, the top 100 most promising genes
were selected.
3. Results
3.1. Towards a comprehensive list of endometriosis-related genes
We run a key word search in the PubMed database for articles relat-
ed to endometriosis and obtained 19,904 articles as a result (from 1980-

Fig. 3. Gene ontology, pathway and network analysis of endometriosis candidate genes. (A) Word cloud of gene ontology (GO) terms enriched among endometriosis candidate genes.
Enrichment test was performed by using BiNGO software configured with GOslim database. The significance cutoff for adjusted p-value was set at 0.01. The font sizes in the word
cloud were proportional to −log10 of adjusted p-value for each enriched GO terms. GO terms were arranged in the biological process (BP) category, the cellular component category
(CC) and the molecular function (MF) category, respectively. (B) Enrichment analysis of pathways for endometriosis candidate genes. The DAVID tool was used and genes are
classified according to the KEGG pathway database. Three pathways were significantly enriched (adjusted p-value b 0.01). (C) The structure of gene network underlying all
endometriosis candidate genes. Nodes represent genes and edges represent gene interactions derived from STRING database. Nodes in red are hub genes with a degree value
exceeding the mean plus two standard deviations.
154 J.-L. Liu, M. Zhao / Genomics 108 (2016) 151–157
 J.-L. Liu, M. Zhao / Genomics 108 (2016) 151–157
Jan to 2016-Jun). Abstracts of these articles were downloaded and proc-
essed through a text mining pipeline (illustrated in Fig. 1A). The number
of articles published on endometriosis is growing exponentially in re-
cent years (Fig. 1B). From these articles, we extracted endometriosis-as-
sociated genes in the title or abstract via text mining. As a result, a total
of 1531 endometriosis-related genes were obtained (Supplementary
Table 1).
We next designed a pipeline to filter the raw text mining results (Fig.
2A). The likelihood that the occurrence for a gene was due to chance
was calculated and then corrected for multiple testing. The results
were used to generate a Manhattan plot (Fig. 2B). We detected 121
genes that reached the statistically significant threshold (adjusted
p b 0.01) (Supplementary Table 2). The top 10 genes were: GNRH1 (go-
nadotropin-releasing hormone 1), MUC16 (mucin 16), FSHB (follicle
stimulating hormone beta polypeptide), CYP19A1 (cytochrome P450
family 19A polypeptide 1), VEGFA (vascular endothelial growth factor
A), PGR (progesterone receptor), ESR1 (estrogen receptor 1), AMH
(anti-Mullerian hormone), PTGS2 (prostaglandin-endoperoxide syn-
thase 2) and HOXA10 (homeobox A10).
3.2. Characterization of endometriosis-related genes
The 121 genes that were significantly associated with endometriosis
were tested for enrichment of functional categories, including gene on-
tology (GO) terms and pathways (Supplementary Table 3). For GO anal-
ysis, all candidate genes were functionally categorized based on GOslim
annotation terms using the BiNGO tool. Enriched GO terms are classified
according to biological process (BP), cellular component (CC) and mo-
lecular function (MF) (Fig. 3A). In the BP category, 19 terms were signif-
icantly enriched, including signal transduction (p = 7.76e − 12), cell
communication (p = 7.76e − 12), multicellular organismal develop-
ment (p = 7.76e − 12), regulation of biological process (p =
8.00e−11), response to external stimulus (p = 2.97e−08), anatomical
structure morphogenesis (p = 5.55e − 08), cell-cell signaling (p =
6.26e − 07), lipid metabolic process (p = 1.44e − 06), behavior (p =
4.40e − 06), reproduction (p = 7.11e − 06), oxygen binding (p =
1.53e − 05), response to stress (p = 8.52e − 05), cell proliferation
(p = 5.12e−04), cell differentiation (p = 1.01e−03), cell death (p =
1.16e − 03), death (p = 1.16e − 03), transcription regulator activity
(p = 2.50e−03), response to endogenous stimulus (p = 4.25e−03),
and response to biotic stimulus (p = 6.96e−03). The enriched CC cat-
egories were extracellular space (p = 2.98e−14), extracellular region
(p = 1.47e−13), proteinaceous extracellular matrix (p = 1.56e−05),
and plasma membrane (p = 3.88e − 03). With respect to molecular
function, the enriched terms were receptor binding (p = 7.76e−12),
protein binding (p = 8.27e − 09), binding (p = 1.03e − 05), signal
transducer activity (p = 1.75e−05), transcription factor activity (p =
1.92e−04), receptor activity (p = 7.82e−04), and calcium ion binding
(p = 8.96e − 03). Pathway analysis was performed using the KEGG
pathway database. Enriched pathways were: cytokine-cytokine recep-
tor interaction (p = 3.62e − 08), steroid hormone biosynthesis (p =
3.04e−06), and apoptosis (p = 9.39e−05) (Fig. 3B).
Additionally, gene network was constructed by using the STRING
software by integrate publicly available interaction data. With the com-
bined score being set to 0.9, we obtained a gene network consisting of
94 nodes connected via 314 edges (Fig. 3C). Topological analysis indi-
cated that the network has some nodes that are highly connected com-
pared to others. These highly connected nodes are known as hub genes.
Using a defined cut-off value, we identified 5 hub genes: TP53 (tumor
protein p53), VEGFA (vascular endothelial growth factor A), AKT1 (v-
Fig. 4. Top 100 novel candidate genes for endometriosis prioritized by the Endeavour software. The full human genome was scanned. Rank ratios for 6 data sources (gene ontology, protein
domains from InterPro, KEGG pathway, microarray expression, gene network from STRING and transcriptional motifs from TRANSFAC) were shown as a heatmap. The integrated rank was
labeled on the left of the heatmap.
155
akt murine thymoma viral oncogene homolog 1), MMP9 (matrix
metallopeptidase 9) and IL6 (interleukin 6). These genes are likely
more important than other genes due to their key positions in the
network.
3.3. Prioritization of novel candidate genes with Endeavour
It has been shown that genes involved in the same disease share ap-
proximately 80% of annotations in gene ontology and InterPro data-
bases. Moreover, genes participated in the same biological pathway
often exhibit a high degree of sequence similarity with other members.
Hence, gene annotations might be indicative of functional importance
for a particular disease. In the present study, we performed novel gene
prioritization by using the Endeavour software to integrate 6 data
sources (gene ontology, protein domains from InterPro, KEGG pathway,
microarray expression, gene network from STRING and transcriptional
motifs from TRANSFAC). The 121 genes that were significantly associat-
ed with endometriosis were used as the training set. The full genome
was scanned and a global ranking for each gene was generated accord-
ing to order statistics. Finally, the top 100 most promising genes were
selected as novel candidate genes for endometriosis (Fig. 4).
Additionally, we also used the raw text mining results (1531 genes)
as the training set and performed gene prioritization by using the En-
deavour software with the same parameters. We used the receiver op-
erating characteristic curves (ROC) to compare the performance of
Endeavour on the two training sets (Fig. 5). The area under the ROC
curve (AUC) statistics served as a useful metric for the performance of
gene prioritization models: whereas an AUC value close to 1 indicates
an excellent gene prioritization model, a curve that lies close to the di-
agonal (AUC = 0.5) has no information content. The AUC values for
the gene prioritization model trained on raw text mining results and
the model trained on filtered results were 0.7322 and 0.8344, respec-
tively. This analysis provides evidence that our filtered data are superior
to raw text mining data.
4. Discussion
Over the past years, the number of articles published on endometri-
osis is growing fast. Numerous genes and pathways have been found to
play a role in this disease. On the other hand, a detailed understanding
of the molecular mechanisms underlying endometriosis is still far
from complete. Under such a situation, the aims of the present study
were to summarize the most reliable gene set associated with endome-
triosis and to predict novel candidate genes that have not been reported
before.
Initially, we performed a systematic analysis of endometriosis-relat-
ed genes using text mining. After manual curation, we identified a total
of 1531 genes. “The wisdom of crowds” refers to the phenomenon in
which the collective knowledge of a community is greater than the
knowledge of any individual [24]. Based on this concept, in the present
study, taking texting results as input, we proposed a gene filtering pro-
cess in a meta-analysis-like manner by computing the likelihood of
finding more than expected occurrences for every gene across the dis-
ease-centered subset of the PubMed database. Our hypothesis is that
those genes most repeatedly mentioned across a large body of publica-
tions on endometriosis might be data-driven causal endometriosis
genes. In this process, we detected 121 genes that reached the statisti-
cally significant threshold (adjusted p b 0.01). These genes may serve
as a highly reliable gene set for endometriosis.
156 J.-L. Liu, M. Zhao / Genomics 108 (2016) 151–157
Fig. 5. Receiver operating characteristic curves (ROC) for gene prioritization results from
the Endeavour software. The blue curve was generated using raw text mining data and
the red curve was generated using filtered data. AUC, area under ROC curve.
Based on gene ontology analysis, a total of 31 terms were significant-
ly enriched among these 121 genes, including signal transduction, cell
communication, cell-cell signaling, lipid metabolic process, response
to stress, cell proliferation, cell differentiation, cell death, extracellular
space, receptor binding, and transcription factor activity. Pathway anal-
ysis revealed that cytokine-cytokine receptor interaction, steroid hor-
mone biosynthesis, and apoptosis were significantly enriched.
In addition, we constructed a gene network by using gene-gene in-
teraction data available in the STRING database. We found that TP53,
VEGFA, AKT1, MMP9 and IL6 were hub genes of this network. The hub
genes are likely more important than other genes due to their key posi-
tions in the network. TP53 gene is a tumor suppressor gene and plays an
important role in the regulation of cell growth and prevention of carci-
nogenesis [25]. TP53 located on chromosome 17 and aneuploidy of this
chromosome had occurred during the development of endometriosis
[26]. A significant decrease of TP53 protein was observed in ovarian le-
sions compared with eutopic endometrium [27]. Additionally, a com-
mon SNP Arg72Pro (rs1042522) has been demonstrated to be
associated with the susceptibility to endometriosis [28–34]. The ectopic
survival of endometrium requires the formation of new blood vessels.
Therefore, angiogenesis is a critical step in developing endometriotic le-
sions [35]. VEGFA, a key angiogenic factor, is increased in the peritoneal
fluid and serum of women with endometriosis [36–38]. Serval studies
regarding VEGF polymorphisms in endometriosis have been performed
[39–44]. The serine-threonine protein kinase AKT1 is highly phosphor-
ylated in eutopic endometrial stromal cells from women with endome-
triosis compared with healthy controls [45,46]. Overactive AKT1 in
endometriotic stromal cells attenuates decidualization through its
downstream target FOXO1 [47]. AKT1 also plays a role in promoting
proliferation and survival of endometriotic cells [48]. The 92-kDa type
IV collagenase MMP9 is a member of secreted zinc metalloproteases
which degrade the collagens of the extracellular matrix [49]. It has
been shown that MMP9 mRNA expression levels were significantly
higher in clinically aggressive pigmented lesions than in normal eutopic
endometrium. Therefore, MMP9 may contribute to survival and inva-
sion of endometriosis [50]. Endometriosis is inflammatory condition as-
sociated with elevated tissue-, peripheral- and peritoneal-cytokines. IL6
is a well-characterized pro-inflammatory cytokine implicated with en-
dometriosis [51]. A recent study has shown that IL6 was increased in en-
dometrial stromal cells isolated from the endometrial biopsies of
women with endometriosis compared to healthy controls [52]. IL6 con-
tributes to pathogenesis of endometriosis through the activation of sig-
nal transducer and activator of transcription (STAT) family of
transcription factors [53]. These data provide clues for the multiple
mechanisms hypothesized to be responsible for the establishment of ec-
topic endometrial tissues, including the migration, implantation, surviv-
al and proliferation of the ectopic endometrial cells.
It has been well established that genes involved in the same disease
usually have a large portion of shared annotations across many data-
bases. Hence, using the reliable 121 genes associated with endometri-
osis as “seed”, we were able to predict novel candidate genes. So far,
several computational methods for gene prioritization have been devel-
oped [54], which are suitable for this task. In the present study, we chose
the Endeavour software, because it is an easy-to-use online tool [55] and
performs best in an unbiased evaluation of similar tools [56]. The En-
deavour tool was configured to integrate 6 data sources: gene ontology,
protein domains from InterPro, KEGG pathway, microarray expression,
gene network from STRING and transcriptional motifs from TRANSFAC.
By scanning the full human genome, we made a list of 100 most prom-
ising candidate genes. These genes deserve further investigation. Addi-
tionally, we also used the raw text mining data (1531 genes) as the
training set and performed gene prioritization by using the Endeavour
software with the same parameters. The outperformance of our filtered
data was noted in term of the area under the ROC curve. This result may
provide validity of our filtering process to select highly reliable genes.
In conclusion, in the present study we generated a highly reliable set
of endometriosis-related genes by statistical test. Characterization of
this gene set was conducted by gene ontology, pathway and network
analysis. This gene set was further used as “seed” to predict novel candi-
date genes. Our study provides in-depth insights into the molecular
mechanism underlying pathogenesis of endometriosis.
Supplementary data to this article can be found online at doi:10.
1016/j.ygeno.2016.10.003.
Conflict of interest
The authors have declared that no competing interests exist.
Acknowledgements
This work was supported by National Natural Science Foundation of
China (grant number 31271602 to Ji-Long Liu) (http://www.nsfc.gov.
cn/).
References
[1] B. Eskenazi, M.L. Warner, Epidemiology of endometriosis, Obstet. Gynecol. Clin. N.
Am. 24 (1997) 235–258.
[2] T.M. D'Hooghe, S. Debrock, Endometriosis, retrograde menstruation and peritoneal
inflammation in women and in baboons, Hum. Reprod. Update 8 (2002) 84–88.
[3] W.P. Dmowski, R.W. Steele, G.F. Baker, Deficient cellular immunity in endometriosis,
Am. J. Obstet. Gynecol. 141 (1981) 377–383.
[4] D.I. Lebovic, M.D. Mueller, R.N. Taylor, Immunobiology of endometriosis, Fertil.
Steril. 75 (2001) 1–10.
[5] C.E. Gargett, K.E. Schwab, J.J. Brosens, P. Puttemans, G. Benagiano, I. Brosens, Poten-
tial role of endometrial stem/progenitor cells in the pathogenesis of early-onset en-
dometriosis, Mol. Hum. Reprod. 20 (2014) 591–598.
[6] R. Grummer, Animal models in endometriosis research, Hum. Reprod. Update 12
(2006) 641–649.
[7] S.A. Treloar, D.T. O'Connor, V.M. O'Connor, N.G. Martin, Genetic influences on endo-
metriosis in an Australian twin sample, Fertil. Steril. 71 (1999) 701–710.
[8] R. Saha, H.J. Pettersson, P. Svedberg, M. Olovsson, A. Bergqvist, L. Marions, P.
Tornvall, R. Kuja-Halkola, Heritability of endometriosis, Fertil. Steril. 104 (2015)
947–952.
[9] N. Rahmioglu, S.A. Missmer, G.W. Montgomery, K.T. Zondervan, Insights into
assessing the genetics of endometriosis, Curr. Obstet. Gynecol. Rep. 1 (2012)
124–137.
 J.-L. Liu, M. Zhao / Genomics 108 (2016) 151–157
[10] N. Rahmioglu, D.R. Nyholt, A.P. Morris, S.A. Missmer, G.W. Montgomery, K.T.
Zondervan, Genetic variants underlying risk of endometriosis: insights from meta-
analysis of eight genome-wide association and replication datasets, Hum. Reprod.
Update 20 (2014) 702–716.
[11] S. Matsuzaki, DNA microarray analysis in endometriosis for development of more
effective targeted therapies, Front. Biosci. (Elite Ed.) 3 (2011) 1139–1153.
[12] C.H. Wei, B.R. Harris, D. Li, T.Z. Berardini, E. Huala, H.Y. Kao, Z. Lu, Accelerating liter-
ature curation with text-mining tools: a case study of using PubTator to curate
genes in PubMed abstracts, Database (Oxford) 2012 (2012), bas041.
[13] C.H. Wei, H.Y. Kao, Z. Lu, PubTator: a web-based text mining tool for assisting
biocuration, Nucleic Acids Res. 41 (2013) W518–W522.
[14] C.H. Wei, H.Y. Kao, Z. Lu, GNormPlus: an integrative approach for tagging genes,
gene families, and protein domains, Biomed. Res. Int. 2015 (2015) 918710.
[15] C.H. Wei, H.Y. Kao, Z. Lu, SR4GN: a species recognition software tool for gene nor-
malization, PLoS One 7 (2012), e38460.
[16] C.H. Wei, H.Y. Kao, Cross-species gene normalization by species inference, BMC
Bioinf., 12 Suppl 8 (2011) S5.
[17] C.H. Wei, R. Leaman, Z. Lu, SimConcept: a hybrid approach for simplifying composite
named entities in biomedicine, ACM BCB 2014 (2014) 138–146.
[18] C.H. Wei, R. Leaman, Z. Lu, SimConcept: a hybrid approach for simplifying composite
named entities in biomedical text, IEEE J. Biomed. Health Inform. 19 (2015)
1385–1391.
[19] S. Sohn, D.C. Comeau, W. Kim, W.J. Wilbur, Abbreviation definition identification
based on automatic precision estimates, BMC Bioinf. 9 (2008) 402.
[20] S. Maere, K. Heymans, M. Kuiper, BiNGO: a Cytoscape plugin to assess overrepresen-
tation of gene ontology categories in biological networks, Bioinformatics 21 (2005)
3448–3449.
[21] D.W. Huang, B.T. Sherman, Q. Tan, J.R. Collins, W.G. Alvord, J. Roayaei, R. Stephens,
M.W. Baseler, H.C. Lane, R.A. Lempicki, The DAVID Gene Functional Classification
Tool: a novel biological module-centric algorithm to functionally analyze large
gene lists, Genome Biol. 8 (2007) R183.
[22] D. Szklarczyk, A. Franceschini, S. Wyder, K. Forslund, D. Heller, J. Huerta-Cepas, M.
Simonovic, A. Roth, A. Santos, K.P. Tsafou, M. Kuhn, P. Bork, L.J. Jensen, C. von
Mering, STRING v10: protein-protein interaction networks, integrated over the
tree of life, Nucleic Acids Res. 43 (2015) D447–D452.
[23] S. Aerts, D. Lambrechts, S. Maity, P. Van Loo, B. Coessens, F. De Smet, L.C.
Tranchevent, B. De Moor, P. Marynen, B. Hassan, P. Carmeliet, Y. Moreau, Gene pri-
oritization through genomic data fusion, Nat. Biotechnol. 24 (2006) 537–544.
[24] J. Surowiecki, The Wisdom of Crowds: Why the Many Are Smarter Than the Few
and How Collective Wisdom Shapes Business, Economies, Societies and Nations,
Knopf Publishing Group, 2005.
[25] S.E. Kern, K.W. Kinzler, S.J. Baker, J.M. Nigro, V. Rotter, A.J. Levine, P. Friedman, C.
Prives, B. Vogelstein, Mutant p53 proteins bind DNA abnormally in vitro, Oncogene
6 (1991) 131–136.
[26] Y. Kosugi, S. Elias, L.R. Malinak, J. Nagata, K. Isaka, M. Takayama, J.L. Simpson, F.Z.
Bischoff, Increased heterogeneity of chromosome 17 aneuploidy in endometriosis,
Am. J. Obstet. Gynecol. 180 (1999) 792–797.
[27] G. Allavena, P. Carrarelli, B. Del Bello, S. Luisi, F. Petraglia, E. Maellaro, Autophagy is
upregulated in ovarian endometriosis: a possible interplay with p53 and heme ox-
ygenase-1, Fertil. Steril. 103 (2015) 1244–1251 (e1241).
[28] M. Ammendola, F. Gloria-Bottini, F. Sesti, E. Piccione, E. Bottini, Association of p53
codon 72 polymorphism with endometriosis, Fertil. Steril. 90 (2008) 406–408.
[29] C.M. Camargo-Kosugi, P. D'Amora, J.P. Kleine, C.V. Carvalho, H. Sato, E. Schor, I.D.
Silva, TP53 gene polymorphisms at codons 11, 72, and 248 and association with en-
dometriosis in a Brazilian population, Genet. Mol. Res. 13 (2014) 6503–6511.
[30] C.C. Chang, Y.Y. Hsieh, F.J. Tsai, C.H. Tsai, H.D. Tsai, C.C. Lin, The proline form of p53
codon 72 polymorphism is associated with endometriosis, Fertil. Steril. 77 (2002)
43–45.
[31] M.P. Gallegos-Arreola, L.E. Figuera-Villanueva, A.M. Puebla-Perez, H. Montoya-
Fuentes, A.E. Suarez-Rincon, G.M. Zuniga-Gonzalez, Association of TP53 gene
codon 72 polymorphism with endometriosis in Mexican women, Genet. Mol. Res.
11 (2012) 1401–1408.
[32] Y.Y. Hsieh, C.S. Lin, P53 codon 11, 72, and 248 gene polymorphisms in endometri-
osis, Int. J. Biol. Sci. 2 (2006) 188–193.
[33] D. Lattuada, P. Vigano, E. Somigliana, A. Abbiati, M. Candiani, A.M. Di Blasio, Analysis
of the codon 72 polymorphism of the TP53 gene in patients with endometriosis,
Mol. Hum. Reprod. 10 (2004) 651–654.
[34] T.H. Ying, C.J. Tseng, S.J. Tsai, S.C. Hsieh, H.Z. Lee, Y.H. Hsieh, D.T. Bau, Association of
p53 and CDKN1A genotypes with endometriosis, Anticancer Res. 31 (2011)
4301–4306.
157
[35] P.G. Groothuis, A.W. Nap, E. Winterhager, R. Grummer, Vascular development in en-
dometriosis, Angiogenesis 8 (2005) 147–156.
[36] W. Kupker, A. Schultze-Mosgau, K. Diedrich, Paracrine changes in the peritoneal en-
vironment of women with endometriosis, Hum. Reprod. Update 4 (1998) 719–723.
[37] J. McLaren, A. Prentice, D.S. Charnock-Jones, S.K. Smith, Vascular endothelial growth
factor (VEGF) concentrations are elevated in peritoneal fluid of women with endo-
metriosis, Hum. Reprod. 11 (1996) 220–223.
[38] V.A. Oliveira, L.G. Abreu, R.A. Ferriani, R.M. Reis, M.D. Moura, Vascular endothelial
growth factor in the plasma, follicular fluid and granulosa cells of women with en-
dometriosis submitted to in vitro fertilization—a pilot study, Gynecol. Endocrinol. 20
(2005) 284–288.
[39] M. Bhanoori, K. Arvind Babu, N.G. Pavankumar Reddy, K. Lakshmi Rao, K. Zondervan,
M. Deenadayal, S. Kennedy, S. Shivaji, The vascular endothelial growth factor (VEGF)
+405GNC 5′-untranslated region polymorphism and increased risk of endometri-
osis in South Indian women: a case control study, Hum. Reprod. 20 (2005)
1844–1849.
[40] R. Cosin, J. Gilabert-Estelles, L.A. Ramon, F. Espana, J. Gilabert, A. Romeu, A. Estelles,
Vascular endothelial growth factor polymorphisms (−460C/T, +405G/C, and 936C/
T) and endometriosis: their influence on vascular endothelial growth factor expres-
sion, Fertil. Steril. 92 (2009) 1214–1220.
[41] D. Gentilini, E. Somigliana, P. Vigano, M. Vignali, M. Busacca, A.M. Di Blasio, The vas-
cular endothelial growth factor +405GNC polymorphism in endometriosis, Hum.
Reprod. 23 (2008) 211–215.
[42] Y. Ikuhashi, S. Yoshida, S. Kennedy, K. Zondervan, N. Takemura, M. Deguchi, N.
Ohara, T. Maruo, Vascular endothelial growth factor +936 C/T polymorphism is as-
sociated with an increased risk of endometriosis in a Japanese population, Acta
Obstet. Gynecol. Scand. 86 (2007) 1352–1358.
[43] S.H. Kim, Y.M. Choi, S.H. Choung, J.K. Jun, J.G. Kim, S.Y. Moon, Vascular endothelial
growth factor gene +405 C/G polymorphism is associated with susceptibility to ad-
vanced stage endometriosis, Hum. Reprod. 20 (2005) 2904–2908.
[44] Z.Z. Zhao, D.R. Nyholt, S. Thomas, S.A. Treloar, G.W. Montgomery, Polymorphisms in
the vascular endothelial growth factor gene and the risk of familial endometriosis,
Mol. Hum. Reprod. 14 (2008) 531–538.
[45] P. Laudanski, R. Charkiewicz, M. Kuzmicki, J. Szamatowicz, J. Swiatecka, B. Mroczko,
J. Niklinski, Profiling of selected angiogenesis-related genes in proliferative eutopic
endometrium of women with endometriosis, Eur. J. Obstet. Gynecol. Reprod. Biol.
172 (2014) 85–92.
[46] P. Laudanski, J. Szamatowicz, O. Kowalczuk, M. Kuzmicki, M. Grabowicz, L.
Chyczewski, Expression of selected tumor suppressor and oncogenes in endometri-
um of women with endometriosis, Hum. Reprod. 24 (2009) 1880–1890.
[47] X. Yin, M.E. Pavone, Z. Lu, J. Wei, J.J. Kim, Increased activation of the PI3K/AKT path-
way compromises decidualization of stromal cells from endometriosis, J. Clin.
Endocrinol. Metab. 97 (2012) E35–E43.
[48] T.H. Kim, Y. Yu, L. Luo, J.P. Lydon, J.W. Jeong, J.J. Kim, Activated AKT pathway pro-
motes establishment of endometriosis, Endocrinology 155 (2014) 1921–1930.
[49] H. Nagase, A.J. Barrett, J.F. Woessner Jr., Nomenclature and glossary of the matrix
metalloproteinases, Matrix 1 (1992) 421–424.
[50] M. Ueda, Y. Yamashita, M. Takehara, Y. Terai, K. Kumagai, K. Ueki, K. Kanda, H.
Yamaguchi, D. Akise, Y.C. Hung, M. Ueki, Survivin gene expression in endometriosis,
J. Clin. Endocrinol. Metab. 87 (2002) 3452–3459.
[51] P.C. Heinrich, I. Behrmann, S. Haan, H.M. Hermanns, G. Muller-Newen, F. Schaper,
Principles of interleukin (IL)-6-type cytokine signalling and its regulation, Biochem.
J. 374 (2003) 1–20.
[52] T. Tsudo, T. Harada, T. Iwabe, M. Tanikawa, Y. Nagano, M. Ito, F. Taniguchi, N.
Terakawa, Altered gene expression and secretion of interleukin-6 in stromal cells
derived from endometriotic tissues, Fertil. Steril. 73 (2000) 205–211.
[53] B.G. Kim, J.Y. Yoo, T.H. Kim, J.H. Shin, J.F. Langenheim, S.D. Ferguson, A.T. Fazleabas,
S.L. Young, B.A. Lessey, J.W. Jeong, Aberrant activation of signal transducer and acti-
vator of transcription-3 (STAT3) signaling in endometriosis, Hum. Reprod. 30
(2015) 1069–1078.
[54] L.C. Tranchevent, F.B. Capdevila, D. Nitsch, B. De Moor, P. De Causmaecker, Y.
Moreau, A guide to web tools to prioritize candidate genes, Brief. Bioinform. 12
(2011) 22–32.
[55] L.C. Tranchevent, R. Barriot, S. Yu, S. Van Vooren, P. Van Loo, B. Coessens, B. De Moor,
S. Aerts, Y. Moreau, ENDEAVOUR update: a web resource for gene prioritization in
multiple species, Nucleic Acids Res. 36 (2008) W377–W384.
[56] D. Bornigen, L.C. Tranchevent, F. Bonachela-Capdevila, K. Devriendt, B. De Moor, P.
De Causmaecker, Y. Moreau, An unbiased evaluation of gene prioritization tools,
Bioinformatics 28 (2012) 3081–3088.