Skip to main content
NCBI home page
AMIA Annual Symposium Proceedings logoLink to AMIA Annual Symposium Proceedings
. 2013 Nov 16;2013:164–173.

Semantic MEDLINE for Discovery Browsing: Using Semantic Predications and the Literature-Based Discovery Paradigm to Elucidate a Mechanism for the Obesity Paradox

Michael J Cairelli 1, Christopher M Miller 1, Marcelo Fiszman 1, T Elizabeth Workman 1, Thomas C Rindflesch 1
PMCID: PMC3900170  PMID: 24551329

Abstract

Applying the principles of literature-based discovery (LBD), we elucidate the paradox that obesity is beneficial in critical care despite contributing to disease generally. Our approach enhances a previous extension to LBD, called “discovery browsing,” and is implemented using Semantic MEDLINE, which summarizes the results of a PubMed search into an interactive graph of semantic predications. The methodology allows a user to construct argumentation underpinning an answer to a biomedical question by engaging the user in an iterative process between system output and user knowledge. Components of the Semantic MEDLINE output graph identified as “interesting” by the user both contribute to subsequent searches and are constructed into a logical chain of relationships constituting an explanatory network in answer to the initial question. Based on this methodology we suggest that phthalates leached from plastic in critical care interventions activate PPAR gamma, which is anti-inflammatory and abundant in obese patients.

Introduction

Our goal is to use literature-based discovery (LBD) to elucidate the paradox that obesity has a beneficial effect in critical care, although it contributes to disease generally. The methodology enhances our previous work1 on “discovery browsing,” which is an extension to LBD that concentrates on elucidating poorly understood areas of biomedicine, rather than necessarily making new discoveries. A key component of “discovery browsing” is “cooperative reciprocity” between human and machine in order to iteratively focus system output to meet the needs of the user. This method of discovery browsing uses Semantic MEDLINE as a framework in which the user exploits semantic predications generated by SemRep2, summarized with topic-relevant schemas3, and visualized as a network of concepts (predication arguments) connected by their relationships (predicates).

Semantic MEDLINE facilitates a novel method, the goal of which is to construct argumentation underpinning an answer to a biomedical question. This argumentation consists of a series of key concepts and their relationships integrated into an overall explanatory network. The method begins with a graphical summary of the citations retrieved from an initial query to Semantic MEDLINE. In “cooperative reciprocity” with these results, the user decides which concepts in the graph should be pursued in the subsequent query to Semantic MEDLINE. This iterative process is continued until the question has been effectively answered.

We illustrate this methodology with results that provide an explanatory mechanism for the obesity paradox. In summary, peroxisome proliferator-activated receptor (PPAR) gamma is most greatly expressed in adipose tissue, activation of PPAR gamma is anti-inflammatory, diethylhexyl phthalate (DEHP) activates PPAR gamma, and DEHP is leached from lines and bags at therapeutic doses in standard interventions in intensive care unit (ICU) patients.

Background

SemRep

SemRep extracts semantic predications from the biomedical research literature2. A predication is a formal representation of textual content that consists of a subject, predicate, and object. Subject and object arguments are concepts from the Unified Medical Language System Metathesaurus4 (UMLS) as accessed through MetaMap5. The predicate is from the Semantic Network6. These predications provide a normalized representation of part of the meaning of the source text in a computable format. As an example, SemRep extracts the predication “phthalate STIMULATES PPAR gamma” from the following sentence:

“We thus performed structural and functional analyses that demonstrate how monoethyl-hexyl-phthalate (MEHP) directly activates PPARgamma and promotes adipogenesis…”

(PMID 17468099)

Semantic MEDLINE

Semantic MEDLINE7 is a web-based system that visualizes SemRep-generated semantic predications extracted from MEDLINE citations stored in SemMedDB8. SemMedDB is a database containing the predications extracted by SemRep from the titles and abstracts of citations in the MEDLINE database8. As of January 31st, 2013, SemMedDB contained 62,776,964 semantic predications from 21,987,674 citations published between January 1st, 1900 and December 31 2012. Predications from retrieved citations are summarized based either on clustered cliques9 or a user-selected schema3. The resulting summary is represented as an interactive graph (Figure 1) with nodes representing arguments and edges representing predicates. Nodes are color-coded by UMLS semantic type and edges are color-coded for individual predicates. Clicking on an edge provides the citation, including the source sentence and PMID, for the represented predication.

Figure 1.

Figure 1.

Open in a new tab

A representative graph produced by Semantic MEDLINE for the search “PPAR gamma and phthalate,” providing specific information for the relationship “phthalate STIMULATES PPAR gamma.”

Literature-based discovery

Literature-based discovery10 is a method of detecting previously unnoticed relationships in the published research literature to provide insight into a poorly understood phenomenon. There are two LBD paradigms. In the first, open discovery, concepts A, B, and C are known and the relationship between A and B and B and C are known, but the relationship between A and C has not been identified. The goal is to discover that relationship. The second paradigm, closed discovery11, is used to explicate an observed phenomenon. A relationship between A and C is assumed but undefined. Relationships between A and B and B and C are identified such that B is provided as a mechanistic link between A and C. Early approaches to automated LBD systems utilized concept co-occurrence12,13,14. Semantic predications were first used for LBD by Hristovski et al15, and more recent work has extended this approach16,17,18,19,20. We have previously used semantic predications in LBD to propose “that cortisol is a mechanistic link connecting declining testosterone levels to age-related alterations in sleep behavior in men.” 21

This work develops the notion of “discovery browsing”1, which extends LBD in that it does not necessarily make a discovery, but elucidates poorly understood phenomena. A key component of discovery browsing is “cooperative reciprocity” between human and machine, in which some formal mechanism provides results from which the user selects the most “interesting/useful.” In the earlier work, the emphasis was on the results generated by a formal mechanism (graph theory)1 while the current proposal focuses on human input.

Serendipitous knowledge discovery

In previous research on using serendipitous knowledge discovery (SKD) for LBD21 we exploited four themes: an iteration of searches, change or clarification of search parameters progressively through the iterations, a seeker’s prior knowledge to provide meaning and relevance to information in search results, and the role of information organization and presentation. In the methodology presented here, information presentation is handled by Semantic MEDLINE, as is the organization of relationships into argument networks. In contrast, SDK addresses “exploring without a specific question.”

The obesity paradox

Although it is well-established that obesity increases morbidity and mortality22, increased obesity has been found to be a predictor of decreased mortality and morbidity in several, principally critical, conditions, including acute coronary syndrome23, heart failure24, stroke25, and respiratory emergencies26, as well as a general ICU phenomenon27. Although explanations range from inaccuracies due to analysis, sampling, or BMI as a measure of obesity28,29,30 to obesity as an inverse of cachexia or COPD31,32, few mechanistic explanations have been posited but include adipose secretion of leptin33 or adiponectin34 causing anti-inflammatory effects and lipophilic sequestration of environmental pollutants35. To date, no mechanistic explanation thoroughly differentiates between the benefits of obesity in acute survival and its detriment over the longer course.

Since we observed that inflammation has unique characteristics in obesity and plays a significant role in the critical conditions demonstrating an obesity paradox, an explanation for the obesity paradox may be gleaned from a combination of knowledge from literature relevant to inflammation and obesity in addition to field-specific journals that describe conditions demonstrating an obesity paradox. In other words, a mechanism might already exist in the literature, although fragmented into different fields.

Methods

Overview

In the cooperative reciprocity that underpins our methodology, Semantic MEDLINE provides system output which the user exploits with domain knowledge to determine the answer to a biomedical question. The process begins with a query (A) to Semantic MEDLINE, based on the question. “Interesting” concepts (B) are identified by hand in the resulting graph and used iteratively36,37 to formulate subsequent queries to the system. Interim results are continually focused and redirected towards answering the initial question.

  • Search1 = A providing Result1 = A → B1,

  • Search2 = A + B1 providing Result2 = B1 → B2,

  • Search3 = A + B1 + B2 providing Result3 = B2 → B3,

  • …

  • Searchn = A1 + B1 + B2 + B3 + … + Bn providing Resultn = Bn−1 → Bn,

A search need not contain all potential search terms but may be a subset of all terms so that Searchn ⊆ {A1, B1, B2, B3, . . ., Bn}. However a search must contain at least one term so Searchn ≠ ∅.

In this paradigm, the B-terms are part of a logical chain of relationships which can be linked together, producing a chain such that A1 → B1 → B2 → B3 → . . . → Bn−1 → Bn. The chain contains a perpetual series of relationships and the terminal relationship is represented by Bn−1 → Bn in place of the standard B → C. Although concepts and relationships can be gathered directly from the predications visualized in the graph, they can also be garnered from reading the source abstracts that the graph links to. As the methodology penetrates iteratively toward a focused resolution of the underlying question of the search these relationships are integrated into an overall explanatory network.

Selecting interestingness concepts

We use several criteria for determining interestingness: 1) uncommon, 2) apparently unrelated, and 3) unfamiliar within the given context. All three depend on prior knowledge of the user and are therefore specific to each individual user. An additional criterion that becomes increasingly important in later iterations of the discovery browsing process is whether a presented relationship connects to a previous B-term or can become the new A-term or the entire argument network. Specific selection of interesting concepts depends on iteration. Uncommon, seemingly unrelated, or unfamiliar concepts are favored in earlier iterations, but as the process continues, relationships with concepts that can connect to arguments already within the chain take precedence.

Using the method to address the obesity paradox

Goal: Discover molecular mechanism contributing to the obesity paradox. Motivation: Approached by clinical researchers after discovering existence of obesity paradox in ICU survival. User: First author. Relevant training: medicine, biochemistry. Status: Clinical researcher with non-expert understanding of endocrinology or PPAR-gamma’s role in metabolism and inflammation.

Iteration 1

Initial search: As shown in Figure 2, the user selects an initial search term that is of a general nature and broadly relevant to the specific research question seeking to be answered. In our example, the initial query to Semantic MEDLINE is “obesity,” resulting in 20,000 citations and 118,325 predications. This search and all subsequent searches were performed on February 4, 2013, limited to publication between January 1, 1900 and December 31, 2012, and limited to a maximum of 20,000 citations.

Figure 2.

Figure 2.

Open in a new tab

Overview of discovery browsing methodology. A general concept relevant to a specific research question is used as an initial search term to enter into Semantic MEDLINE. Interesting concepts are determined in the graph and included in subsequent searches. As iterations continue, interesting relationships are identified which eventually find a position in the logical chain that provides a solution to the initial research question. When this chain is represented in graphical form, it becomes the argument network.

Selection of interesting results: Semantic MEDLINE produced a summary graph containing fewer than 50 concepts, most of which were well known regarding obesity (e.g., Adipocytes, Diabetes) or non-specific (e.g., Body tissue, Physical activity). PPAR gamma (a receptor significant in obesity) was one of a small number of concepts that were candidates for further iterative searching. A provided citation (PMID: 21488194) noted PPAR gamma’s role in regulation of inflammatory response, and inflammation was known to be significant for all of the conditions exhibiting an obesity paradox.

Iteration 2

Composite search: The interesting concept “PPAR gamma” was combined with the initial query to become “obesity and PPAR gamma.” This search returned 1346 citations and 13,224 predications, yielding 6733 predications after summarization.

Selection of interesting results: Results from the second search include the predications “Adipose tissue LOCATION_OF PPAR gamma” (PMID 8647948), which is a well-known relationship and therefore less interesting but included within our logical chain for its connectivity within the logical chain. An additional predication of note was found within the graph: “phthalate STIMULATES PPAR gamma” (PMID 22017230). The concept “phthalate” was deemed interesting because it was uncommon, apparently unrelated, and unfamiliar to the user in this or any other context.

Iterations 3

Composite search: The interesting concept from Iteration 2, “phthalate,” is combined with the previous interesting concept “PPAR gamma” from Iteration 1 to attempt to expand the discourse on the role of phthalates with this receptor known to be centrally positioned in obesity. The search becomes “PPAR gamma and phthalate.” This search returns 32 citations and 368 predications which are reduced to 135 predications after summarization.

Selection of interesting results: This graph produced contains “phthalates INTERACTS_WITH PPAR gamma” (PMID 17468099) again and this abstract specifies that MEHP activates PPAR gamma, which we include as “MEHP STIMULATES PPAR gamma.” The source abstract from a second, similar predication, “Mono(2-ethylhexyl) phthalate STIMULATES Peroxisome Proliferator-Activated Receptors” (PMID 16326050) provided the relationship “DEHP METABOLIZES_TO MEHP.”

Iterations 4

Composite search: To explore the significance of MEHP in the critical care population it is combined with concepts that include this group including the productive query “mono-(2-ethylhexyl) phthalate and intensive care unit.” 7 citations with 51 predications were returned which were summarized to 6 predications by Semantic MEDLINE.

Selection of interesting results: This iteration provides the enticing predication “Blood LOCATION_OF Diethylhexyl Phthalate” (PMID 14594632) with additional information in the abstract that Diethylhexyl Phthalate (DEHP) is the most common plasticizer (i.e., used for flexibility and clarity) in polyvinylchloride and present in blood due to environmental exposure. We summarize these relationships as “PVC CONTAINS DEHP” and “ICU interventions INCREASE PVC exposure,” knowing the common use of PVC in bags and lines in a medical setting.

Iterations 5

Composite search: To further investigate the possibility of PPAR gamma’s role in inflammation being of significance in the conditions displaying an obesity paradox the receptor was also combined with “intensive care unit” analogous to Iteration 4: “PPAR gamma and intensive care unit.” The 150 predications from the 12 citations returned were reduced to 28 predications after summarization.

Selection of interesting results: This search produced a graph containing “PPAR gamma ASSOCIATED_WITH Injury” (PMID 17101152), and the abstract clarified the association by demonstrating that down regulation of PPARG is followed by a proinflammatory response subsequent to sepsis, which is one of the conditions exhibiting the obesity paradox. The proinflammatory effect of down regulation of PPAR gamma suggests that PPAR gamma has an anti-inflammatory role, which we summarized in the relationship “PPAR gamma DECREASES Inflammation.”

Linking relationships

The relationships garnered from the searches in the 5 iterations outlined where combined into a logical chain. This includes relationships directly provided in the graphs as semantic predications as well as relationships provided in the source abstracts, using more specificity for the concepts and/or their relationship. For example, in Iteration 2 the predication “phthalate STIMULATES PPAR gamma” was further specified in the abstract by defining the phthalate specifically as MEHP (PMID 22017230).

Results

Individual relationships

As listed below, seven individual relationships were selected and included in the final logical chain in response to the initial search question. Italicized relationships indicate inclusion of information from abstracts, bold text represents commonly known relationships, whereas those that are neither italicized or bold were predications provided in Semantic MEDLINE graphs.

  1. DEHP METABOLIZES_TO MEHP

  2. MEHP STIMULATES PPAR gamma

  3. PPAR gamma DECREASES inflammation

  4. Inflammation INCREASES mortality and morbidity

  5. ICU interventions INCREASE PVC exposure

  6. PVC CONTAINS DEHP

  7. Obesity INCREASES Adipose tissue

  8. Adipose tissue LOCATION_OF PPAR gamma

Logical chain and argument network

Our argument network is composed of the three logical chains listed below. The main logical chain (1) provides a link from DEHP exposure to decreased morbidity and mortality mediated through PPAR gamma. One auxiliary chain (2) describes an increase of exposure to PVC and therefore MEHP with ICU interventions and the other (3) suggests an increase in PPAR gamma activity in the obese due to an increase in adipose tissue.

  1. DEHP METABOLIZED_TO MEHP STIMULATES PPAR gamma DECREASES inflammation INCREASES mortality/morbidity

  2. ICU interventions INCREASE PVC LOCATION_OF DEHP

  3. Obesity INCREASES Adipose tissue LOCATION_OF PPAR gamma

Discussion

To investigate whether this discovery was novel we submitted a series of PubMed searches to ascertain whether a link between DEHP and the obesity paradox had ever been suggested. A search for “(phthalate OR DEHP OR MEHP) AND obesity” returned 28 citations. The citations were almost all concerned with phthalates inducing obesity with prolonged environmental exposure with a few focused more narrowly on exposure levels within this context and a single citation focused on food safety so also concerned with phthalate toxicity. Searching for “(phthalate OR DEHP OR MEHP) AND obesity paradox” returned no results, suggesting that within MEDLINE no articles linked the concept of phthalates with the obesity paradox. It is worth noting that the obesity paradox is not a large topic of conversation within the biomedical literature as the query “obesity paradox” produced only 220 articles.

Revisiting the A – B – C paradigm

This discovery browsing methodology extends the traditional A-B-C paradigm of LBD by arranging sets of relationships into logical chains and facilitates arranging the logical chains into an argument network. It is worth revisiting the three logical chains included in our argument network in the context of the traditional A-B-C approach, to see how the individual branches are represented in this traditional notation. The longest chain can be summarized as the role of DEHP in mortality and morbidity. As can be seen in Table 1, this chain consists of a series of A – B and B – C components being combined into A – C relationships, requiring three rounds to move through four original relationships. Each line results in a combined (A – C) relationship which is ported to the A – B relationship of the next line. Ultimately, this first logical chain is summarized as “DEHP DECREASES mortality and morbidity.” The second logical chain describes the specificity of the initial logical chain to the obesity paradox or “ICU” conditions. Since there are only two component relationships, they are combined in one round to “ICU interventions INCREASE DEHP exposure.” The third chain describing the specificity of the main chain to obese patients is also completed in a single round, yielding “obesity INCREASES PPAR gamma.”

Table 1.

A–B, B–C, and resulting A–C relationships for each logical chain.

A – B B – C A – C
Role of DEHP in mortality and morbidity
DEHP METABOLIZES_TO MEHP MEHP STIMULATES PPAR gamma DEHP STIMULATES PPAR gamma
DEHP STIMULATES PPAR gamma PPAR gamma DECREASES inflammation DEHP DECREASES Inflammation
DEHP DECREASES Inflammation Inflammation INCREASES mortality and morbidity DEHP DECREASES mortality and morbidity
Specificity to ICU
ICU interventions INCREASE PVC exposure PVC LOCATION_OF DEHP ICU interventions INCREASE DEHP exposure
Specificity to obese patients
Obesity INCREASES Adipose tissue Adipose tissue LOCATION_OF PPAR gamma obesity INCREASES PPAR gamma
Open in a new tab

PPAR gamma’s anti-inflammatory role

Although initially studied for its role in adipogenesis, a more recent study of PPAR gamma has alluded to its significant anti-inflammatory activity (Szanto, 2008). An abundance of studies have outlined anti-inflammatory effects of PPAR gamma activation in animal models and cell culture, which include reduction of inflammatory effects in macrophages, dendritic cells, and macrophages, reduction of airway inflammation related to asthma and COPD, neural and cardiac autoimmunity, colitis, arthritis, and dermatitis among others38. These studies have provided sufficient evidence to warrant trials of PPAR gamma agonists in high inflammatory states, including cardiovascular diseases such as myocardial infarction, acute coronary syndrome, stroke, and carotid plaque formation39,40,41, chronic kidney disease42, ulcerative colitis43, and malaria44. Of these, all but the final two have demonstrated an obesity paradox.

DEHP exposure in obesity paradox interventions

Although the historical focus of the bioactivity of DEHP has been on possible carcinogenesis45 and disturbance of male reproductive tract development46, the mechanism provided by discovery browsing with Semantic MEDLINE suggests it may have a positive effect on obese patients undergoing critical care interventions. However, due to the concern of DEHP having potential toxic effects through PVC exposure, the FDA conducted a study detailing the exposure ranges for various interventions, determining that the exposure levels did not put adults at reasonable risk for toxic effects47. They do, however, provide support for the idea that many procedures common in obesity paradox conditions provide a dosage that is likely to elucidate the anti-inflammatory effects of PPAR gamma agonism. For instance, blood transfusion, TPN administration, coronary artery bypass graft, hemodialysis, enteral nutrition, and certain IV infusion conditions provide doses that may be significant. These doses are exponentially lower than the cancer studies45, but much closer to those demonstrating similar agonistic activity of DEHP to commonly prescribed PPAR gamma agonists48.

Obesity increases PPAR gamma activity

Our argument network suggests that because PPAR gamma is expressed most in adipose tissue, and obesity equates with a significant increase in adiposity, there should be a corresponding increase in PPAR gamma. Although rodent studies have demonstrated an increase of PPAR gamma expression in adipose and diaphragm49, bone marrow and splenocytes50, and lung51; only one study has found increased expression of PPAR-gamma in humans, specifically in the liver of obese patients with non-alcoholic fatty liver52. This lack of evidence for an increase in gene expression is coupled with a complete lack of protein-level data, so there is no direct evidence for a general increase of PPAR gamma abundance in obesity. However, a recent study demonstrates the same effect in a mouse model of acute pancreatitis, showing that obese mice treated with the PPAR gamma agonist rosiglitazone have decreased disease severity and decreased lethality compared to non-obese mice after induced pancreatitis, which is unexpected since obesity usually increases pancreatitis complications53.

Advantage of Semantic MEDLINE over standard information retrieval

Semantic MEDLINE provides a summary of search results that reduces the burden of information processing on the user. An equivalent search for “obesity” using PubMed returned 173,875 citations. The first mention of PPAR gamma is in the 20th citation but includes no mention of inflammation. The 99th citation is the first to mention both concepts, but PPAR gamma is mentioned in the context of a marker for obesity and not directly connected to inflammation. Finally, citation 517 postulates that the PPAR gamma agonist, rosiglitazone, may decrease inflammation.

Conclusion

We utilized the principles of literature-based discovery (LBD) to elucidate the paradox that obesity has a beneficial effect in critical care, although it is well-known to contribute to disease generally. This approach enhances our previous work on “discovery browsing,” which is an extension to LBD and relies on semantic predications extracted by SemRep from all of MEDLINE. It is implemented with the Semantic MEDLINE web application, which summarizes predications extracted from the results of PubMed searches and presents results in interactive graphs. The methodology allows a user to construct argumentation underpinning an answer to a biomedical question. In response to a biomedical question, a user engages in an iterative process of cooperative agreement between Semantic MEDLINE output and the user’s own background knowledge. Components of the Semantic MEDLINE output graph identified as “interesting” by the user both contribute to subsequent searches and are constructed into a logical chain of relationships constituting an explanatory network in answer to the initial question. Based on this methodology we suggest that phthalates leached from plastic in the intensive care unit activate PPAR gamma, which is anti-inflammatory and is abundant in obese patients.

Figure 3.

Figure 3.

Open in a new tab

Argument network for the obesity paradox. The main logical chain (vertical) describes the role of DEHP in reducing mortality and morbidity and context is provided by the auxiliary chains detailing the increase of DEHP exposure in the ICU and an increase in PPAR gamma in obesity.

Acknowledgments

This research was supported in part by an appointment to the NLM Research Participation Program. This program is administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the U.S. Department of Energy and the National Library of Medicine. This research was also supported in part by the Intra mural Research Program of the NIH, National Library of Medicine.

References

  • 1.Wilkowski B, Fiszman M, Miller CM, Hristovski D, Arabandi S, Rosemblat G, Rindflesch TC. Graph-based methods for discovery browsing with semantic predications. AMIA Annu Symp Proc. 2011;2011:1514–23. [PMC free article] [PubMed] [Google Scholar]
  • 2.Rindflesch TC, Fiszman M. The interaction of domain knowledge and linguistic structure in natural language processing: interpreting hypernymic propositions in biomedical text. J Biomed Inform. 2003;36(6):462–77. doi: 10.1016/j.jbi.2003.11.003. [DOI] [PubMed] [Google Scholar]
  • 3.Fiszman M, Rindflesch TC, Kilicoglu H. Summarization of an online medical encyclopedia. Stud Health Technol Inform. 2004;107(Pt 1):506–10. [PubMed] [Google Scholar]
  • 4.Bodenreider O. The Unified Medical Language System (UMLS): integrating biomedical terminology. Nucleic Acids Res. 2004;32(Database issue):D267–70. doi: 10.1093/nar/gkh061. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Aronson AR, Lang FM. An overview of MetaMap: historical perspective and recent advances. J Am Med Inform Assoc. 2010;17(3):229–36. doi: 10.1136/jamia.2009.002733. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Lindberg DA, Humphreys BL, McCray AT. The Unified Medical Language System. Methods of Information in Medicine. 1993;32(4):281–291. doi: 10.1055/s-0038-1634945. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Kilicoglu H, Fiszman M, Rodriguez A, Shin D, Ripple A, Rindflesch T. Semantic MEDLINE: A Web Application to Manage the Results of PubMed Searches. Third International Symposium on Semantic Mining in Biomedicine; 2008. [Google Scholar]
  • 8.Kilicoglu H, Shin D, Fiszman M, Rosemblat G, Rindflesch TC. SemMedDB: a PubMed-scale repository of biomedical semantic predications. Bioinformatics. 2012;28(23):3158–60. doi: 10.1093/bioinformatics/bts591. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Zhang H, Fiszman M, Shin D, Miller CM, Rosemblat G, Rindflesch TC. Degree centrality for semantic abstraction summarization of therapeutic studies. J Biomed Inform. 2011;44(5):830–8. doi: 10.1016/j.jbi.2011.05.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Swanson DR. Fish oil, Raynaud’s syndrome, and undiscovered public knowledge. Perspectives in Biology and Medicine. 1986;30(1):7–18. doi: 10.1353/pbm.1986.0087. [DOI] [PubMed] [Google Scholar]
  • 11.Swanson DR. Migraine and magnesium: eleven neglected connections. Perspectives in Biology and Medicine. 1988;31(4):526–557. doi: 10.1353/pbm.1988.0009. [DOI] [PubMed] [Google Scholar]
  • 12.Swanson DR, Smalheiser NR. An interactive system for finding complementary literatures: a stimulus to scientific discovery. Artificial Intelligence. 1997;91(2):183–203. [Google Scholar]
  • 13.Hristovski D, Stare J, Peterlin B, Dzeroski S. Studies in Health Technology and Informatics. IOS Press; 2001. Supporting discovery in medicine by association rule mining in Medline and UMLS; pp. 1344–1348. [PubMed] [Google Scholar]
  • 14.Weeber M, Kors JA, Mons B. Online tools to support literature-based discovery in the life sciences. Briefings in Bioinformatics. 2005;6(3):277–286. doi: 10.1093/bib/6.3.277. [DOI] [PubMed] [Google Scholar]
  • 15.Hristovski D, Friedman C, Rindflesch TC, Peterlin B. Exploiting semantic relations for literature-based discovery. AMIA Annual Symposium Proceedings; 2006. 2006. pp. 349–353. [PMC free article] [PubMed] [Google Scholar]
  • 16.Ahlers CB, Hristovski D, Kilicoglu H, Rindflesch TC. Using the literature-based discovery paradigm to investigate drug mechanisms. AMIA Annual Symposium Proceedings; 2007. pp. 6–10. [PMC free article] [PubMed] [Google Scholar]
  • 17.Cohen T, Whitfield GK, Schvaneveldt RW, Mukund K, Rindflesch T. EpiphaNet: an interactive tool to support biomedical discoveries. Journal of Biomedical Discovery and Collaboration. 2010;5:22–44. [PMC free article] [PubMed] [Google Scholar]
  • 18.Hristovski D, Kastrin A, Peterlin B, Rindflesch T. Linking Literature, Information, and Knowledge for Biology. 2010. Combining semantic relations and DNA microarray data for novel hypotheses generation; pp. 53–61. [Google Scholar]
  • 19.Goodwin JC, Cohen T, Rindflesch TC. Discovery by scent: Discovery browsing system based on the Information Foraging Theory. IEEE International Conference on Bioinformatics and Biomedicine Workshops; 2012. [Google Scholar]
  • 20.Workman T, Fiszman M, Rindflesch TC, Nahl D. Framing Serendipitous Information Seeking Behavior for Facilitating Literature-Based Discovery: A Proposed Model 2013. (In press)
  • 21.Miller CM, Rindflesch TC, Fiszman M, Hristovski D, Shin D, Rosemblat G, Zhang H, Strohl KP. A closed literature-based discovery technique finds a mechanistic link between hypogonadism and diminished sleep quality in aging men. Sleep. 2012;35(2):279–85. doi: 10.5665/sleep.1640. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Calle EE, Thun MJ, Petrelli JM, Rodriguez C, Heath CW., Jr Body-mass index and mortality in a prospective cohort of U.S. adults. N Engl J Med. 1999;341(15):1097–105. doi: 10.1056/NEJM199910073411501. [DOI] [PubMed] [Google Scholar]
  • 23.Lavie CJ, Milani RV, Ventura HO. Impact of obesity on outcomes in myocardial infarction combating the “obesity paradox”. J Am Coll Cardiol. 2011;58(25):2651–3. doi: 10.1016/j.jacc.2011.09.032. [DOI] [PubMed] [Google Scholar]
  • 24.Arena R, Myers J, Abella J, Pinkstaff S, Brubaker P, Moore B, Kitzman D, Peberdy MA, Bensimhon D, Chase P, Forman D, West E, Guazzi M. Influence of etiology of heart failure on the obesity paradox. Am J Cardiol. 2009;104(8):1116–21. doi: 10.1016/j.amjcard.2009.05.059. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Katsnelson M, Rundek T. Obesity paradox and stroke: noticing the (fat) man behind the curtain. Stroke. 2011;42(12):3331–2. doi: 10.1161/STROKEAHA.111.632471. [DOI] [PubMed] [Google Scholar]
  • 26.Memtsoudis SG, Bombardieri AM, Ma Y, Walz JM, Chiu YL, Mazumdar M. Mortality of patients with respiratory insufficiency and adult respiratory distress syndrome after surgery: the obesity paradox. J Intensive Care Med. 2012;27(5):306–11. doi: 10.1177/0885066611411410. [DOI] [PubMed] [Google Scholar]
  • 27.Abhyankar S, Leishear K, Callaghan FM, Demner-Fushman D, McDonald CJ. Lower short- and long-term mortality associated with overweight and obesity in a large cohort study of adult intensive care unit patients. Crit Care. 2012;16(6):R235. doi: 10.1186/cc11903. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Frankenstein L, Zugck C, Nelles M, Schellberg D, Katus HA, Remppis BA. The obesity paradox in stable chronic heart failure does not persist after matching for indicators of disease severity and confounders. Eur J Heart Fail. 2009;11(12):1189–94. doi: 10.1093/eurjhf/hfp150. [DOI] [PubMed] [Google Scholar]
  • 29.Bray GA. Risk factors: The obesity paradox-an artifact of small sample size? Nat Rev Cardiol. 2009;6(9):561–2. doi: 10.1038/nrcardio.2009.140. [DOI] [PubMed] [Google Scholar]
  • 30.Agarwal R, Bills JE, Light RP. Diagnosing obesity by body mass index in chronic kidney disease: an explanation for the “obesity paradox?”. Hypertension. 2010;56(5):893–900. doi: 10.1161/HYPERTENSIONAHA.110.160747. [DOI] [PubMed] [Google Scholar]
  • 31.Freeman LM. The pathophysiology of cardiac cachexia. Curr Opin Support Palliat Care. 2009;3(4):276–81. doi: 10.1097/SPC.0b013e32833237f1. [DOI] [PubMed] [Google Scholar]
  • 32.Galal W, van Gestel YR, Hoeks SE, Sin DD, Winkel TA, Bax JJ, Verhagen H, Awara AM, Klein J, van Domburg RT, Poldermans D. The obesity paradox in patients with peripheral arterial disease. Chest. 2008;134(5):925–30. doi: 10.1378/chest.08-0418. [DOI] [PubMed] [Google Scholar]
  • 33.Sweeney G. Cardiovascular effects of leptin. Nat Rev Cardiol. 2010;7(1):22–9. doi: 10.1038/nrcardio.2009.224. [DOI] [PubMed] [Google Scholar]
  • 34.Tankó LB, Christiansen C. Can the obesity paradox be explained by the protective effects of peripheral adiposity? Arch Intern Med. 2005;165(15):1796–7. doi: 10.1001/archinte.165.15.1796-b. [DOI] [PubMed] [Google Scholar]
  • 35.Hong NS, Kim KS, Lee IK, Lind PM, Lind L, Jacobs DR, Lee DH. The association between obesity and mortality in the elderly differs by serum concentrations of persistent organic pollutants: a possible explanation for the obesity paradox. Int J Obes (Lond) 2012;36(9):1170–5. doi: 10.1038/ijo.2011.187. [DOI] [PubMed] [Google Scholar]
  • 36.Salton G. Automatic Information Organization and Retrieval. New York: McGraw-Hill; 1968. [Google Scholar]
  • 37.Bates MJ. The Design of Browsing and Berrypicking Techniques for the Online Search Interface. Online Review. 1989;13(5):407–424. [Google Scholar]
  • 38.Szanto A, Nagy L. The many faces of PPARgamma: anti-inflammatory by any means? Immunobiology. 2008;213(9–10):789–803. doi: 10.1016/j.imbio.2008.07.015. [DOI] [PubMed] [Google Scholar]
  • 39.Erdmann E, Dormandy JA, Charbonnel B, Massi-Benedetti M, Moules IK, Skene AM, PROactive Investigators The effect of pioglitazone on recurrent myocardial infarction in 2,445 patients with type 2 diabetes and previous myocardial infarction: results from the PROactive (PROactive 05) Study. J Am Coll Cardiol. 2007;49(17):1772–80. doi: 10.1016/j.jacc.2006.12.048. [DOI] [PubMed] [Google Scholar]
  • 40.Wilcox R, Kupfer S, Erdmann E, PROactive Study investigators Effects of pioglitazone on major adverse cardiovascular events in high-risk patients with type 2 diabetes: results from PROspective pioglitAzone Clinical Trial In macro Vascular Events (PROactive 10) Am Heart J. 2008;155(4):712–7. doi: 10.1016/j.ahj.2007.11.029. [DOI] [PubMed] [Google Scholar]
  • 41.Hirano M, Nakamura T, Kitta Y, Yano T, Kobayashi T, Sano K, Fujioka D, Saito Y, Kodama Y, Kawabata K, Nakamura K, Obata JE, Kugiyama K. Rapid improvement of carotid plaque echogenicity within 1 month of pioglitazone treatment in patients with acute coronary syndrome. Atherosclerosis. 2009;203(2):483–8. doi: 10.1016/j.atherosclerosis.2008.07.023. [DOI] [PubMed] [Google Scholar]
  • 42.Schneider CA, Ferrannini E, Defronzo R, Schernthaner G, Yates J, Erdmann E. Effect of pioglitazone on cardiovascular outcome in diabetes and chronic kidney disease. J Am Soc Nephrol. 2008;19(1):182–7. doi: 10.1681/ASN.2007060678. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Liang HL, Ouyang Q. [A clinical trial of rosiglitazone and 5-aminosalicylate combination for ulcerative colitis] Zhonghua Nei Ke Za Zhi. 2006;45(7):548–51. [PubMed] [Google Scholar]
  • 44.Boggild AK, Krudsood S, Patel SN, Serghides L, Tangpukdee N, Katz K, Wilairatana P, Liles WC, Looareesuwan S, Kain KC. Use of peroxisome proliferator-activated receptor gamma agonists as adjunctive treatment for Plasmodium falciparum malaria: a randomized, double-blind, placebo-controlled trial. Clin Infect Dis. 2009;49(6):841–9. doi: 10.1086/605431. [DOI] [PubMed] [Google Scholar]
  • 45.Beliles R, Salinas JA, Kluwe WM. A review of di(2-ethylhexyl)phthalate (DEHP) risk assessments. Drug Metab Rev. 1989;21(1):3–12. doi: 10.3109/03602538909029952. [DOI] [PubMed] [Google Scholar]
  • 46.Latini G, De Felice C, Verrotti A. Plasticizers, infant nutrition and reproductive health. Reprod Toxicol. 2004 Nov;19(1):27–33. doi: 10.1016/j.reprotox.2004.05.011. [DOI] [PubMed] [Google Scholar]
  • 47.U.S. Food and Drug Administration . Safety Assessment of Released from PVC Medical Devices. Rockville, MD: 2002. p. 47. [Google Scholar]
  • 48.Feige JN, Gelman L, Rossi D, Zoete V, Métivier R, Tudor C, Anghel SI, Grosdidier A, Lathion C, Engelborghs Y, Michielin O, Wahli W, Desvergne B. The endocrine disruptor monoethyl-hexyl-phthalate is a selective peroxisome proliferator-activated receptor gamma modulator that promotes adipogenesis. J Biol Chem. 2007;282(26):19152–66. doi: 10.1074/jbc.M702724200. [DOI] [PubMed] [Google Scholar]
  • 49.Gorla-Bajszczak A, Siegrist-Kaiser C, Boss O, Burger AG, Meier CA. Expression of peroxisome proliferator-activated receptors in lean and obese Zucker rats. Eur J Endocrinol. 2000;142(1):71–8. doi: 10.1530/eje.0.1420071. [DOI] [PubMed] [Google Scholar]
  • 50.Halade GV, Rahman MM, Williams PJ, Fernandes G. High fat diet-induced animal model of age-associated obesity and osteoporosis. J Nutr Biochem. 2010;21(12):1162–9. doi: 10.1016/j.jnutbio.2009.10.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Hwang SJ, Kim JH, Shim JW, Kim DS, Jung HL, Park MS, Lee WY, Kim SY, Shim JY. Peroxisome proliferator-activated receptor-gamma expression in the lung tissue of obese rats. Yonsei Med J. 2011;52(3):495–501. doi: 10.3349/ymj.2011.52.3.495. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Pettinelli P, Videla LA. Up-regulation of PPAR-gamma mRNA expression in the liver of obese patients: an additional reinforcing lipogenic mechanism to SREBP-1c induction. J Clin Endocrinol Metab. 2011 May;96(5):1424–30. doi: 10.1210/jc.2010-2129. [DOI] [PubMed] [Google Scholar]
  • 53.Pini M, Rhodes DH, Castellanos KJ, Cabay RJ, Grady EF, Fantuzzi G. Rosiglitazone improves survival and hastens recovery from pancreatic inflammation in obese mice. PLoS One. 2012;7(7):e40944. doi: 10.1371/journal.pone.0040944. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from AMIA Annual Symposium Proceedings are provided here courtesy of American Medical Informatics Association

RESOURCES