Optimization and Applications of Large-Scale Biomedical Event Networks

The overwhelming amount and unprecedented speed of publication in the biomedical domain make it difficult for life science researchers to acquire and maintain a broad view of the field and gather all information that would be relevant for their research. As a response to this problem, the BioNLP (Biomedical Natural Language Processing) community of researches has emerged and strives to assist life science researchers by developing modern natural language processing (NLP), information extraction (IE) and information retrieval (IR) methods that can be applied at large-scale, to scan the whole publicly available biomedical literature and extract and aggregate the information found within, while automatically normalizing the variability of natural language statements. Among different tasks, biomedical event extraction has received much attention within BioNLP community recently. Biomedical event extraction constitutes the identification of biological processes and interactions described in biomedical literature, and their representation as a set of recursive event structures. The 2009–2013 series of BioNLP Shared Tasks on Event Extraction have given raise to a number of event extraction systems, several of which have been applied at a large scale (the full set of PubMed abstracts and PubMed Central Open Access full text articles), leading to creation of massive biomedical event databases, each of which containing millions of events. Sinece top-ranking event extraction systems are based on machine-learning approach and are trained on the narrow-domain, carefully selected Shared Task training data, their performance drops when being faced with the topically highly varied PubMed and PubMed Central documents. Specifically, false-positive predictions by these systems lead to generation of incorrect biomolecular events which are spotted by the end-users. This thesis proposes a novel post-processing approach, utilizing a combination of supervised and unsupervised learning techniques, that can automatically identify and filter out a considerable proportion of incorrect events from large-scale event databases, thus increasing the general credibility of those databases. The second part of this thesis is dedicated to a system we developed for hypothesis generation from large-scale event databases, which is able to discover novel biomolecular interactions among genes/gene-products. We cast the hypothesis generation problem as a supervised network topology prediction, i.e predicting new edges in the network, as well as types and directions for these edges, utilizing a set of features that can be extracted from large biomedical event networks. Routine machine learning evaluation results, as well as manual evaluation results suggest that the problem is indeed learnable. This work won the Best Paper Award in The 5th International Symposium on Languages in Biology and Medicine (LBM 2013).

A Visual Analytics Approach to Comparing Cohorts of Event Sequences

Sequences of timestamped events are currently being generated across nearly every domain of data analytics, from e-commerce web logging to electronic health records used by doctors and medical researchers. Every day, this data type is reviewed by humans who apply statistical tests, hoping to learn everything they can about how these processes work, why they break, and how they can be improved upon. To further uncover how these processes work the way they do, researchers often compare two groups, or cohorts, of event sequences to find the differences and similarities between outcomes and processes. With temporal event sequence data, this task is complex because of the variety of ways single events and sequences of events can differ between the two cohorts of records: the structure of the event sequences (e.g., event order, co-occurring events, or frequencies of events), the attributes about the events and records (e.g., gender of a patient), or metrics about the timestamps themselves (e.g., duration of an event). Running statistical tests to cover all these cases and determining which results are significant becomes cumbersome. Current visual analytics tools for comparing groups of event sequences emphasize a purely statistical or purely visual approach for comparison. Visual analytics tools leverage humans' ability to easily see patterns and anomalies that they were not expecting, but is limited by uncertainty in findings. Statistical tools emphasize finding significant differences in the data, but often requires researchers have a concrete question and doesn't facilitate more general exploration of the data. Combining visual analytics tools with statistical methods leverages the benefits of both approaches for quicker and easier insight discovery. Integrating statistics into a visualization tool presents many challenges on the frontend (e.g., displaying the results of many different metrics concisely) and in the backend (e.g., scalability challenges with running various metrics on multi-dimensional data at once). I begin by exploring the problem of comparing cohorts of event sequences and understanding the questions that analysts commonly ask in this task. From there, I demonstrate that combining automated statistics with an interactive user interface amplifies the benefits of both types of tools, thereby enabling analysts to conduct quicker and easier data exploration, hypothesis generation, and insight discovery. The direct contributions of this dissertation are: (1) a taxonomy of metrics for comparing cohorts of temporal event sequences, (2) a statistical framework for exploratory data analysis with a method I refer to as high-volume hypothesis testing (HVHT), (3) a family of visualizations and guidelines for interaction techniques that are useful for understanding and parsing the results, and (4) a user study, five long-term case studies, and five short-term case studies which demonstrate the utility and impact of these methods in various domains: four in the medical domain, one in web log analysis, two in education, and one each in social networks, sports analytics, and security. My dissertation contributes an understanding of how cohorts of temporal event sequences are commonly compared and the difficulties associated with applying and parsing the results of these metrics. It also contributes a set of visualizations, algorithms, and design guidelines for balancing automated statistics with user-driven analysis to guide users to significant, distinguishing features between cohorts. This work opens avenues for future research in comparing two or more groups of temporal event sequences, opening traditional machine learning and data mining techniques to user interaction, and extending the principles found in this dissertation to data types beyond temporal event sequences.

Testing the Prebish-Singer hypothesis using second-generation panel data stationarity tests with a break

In this paper, we test the Prebish-Singer (PS) hypothesis, which states that real commodity prices decline in the long run, using two recent powerful panel data stationarity tests accounting for cross-sectional dependence and a structural break. We find that the hypothesis cannot be rejected for most commodities other than oil.

Next Generation SHRM Research: From Covariation to Causation

[Excerpt] In response to the longstanding and repeated criticisms that HR does not add value to organizations, the past 10 years has seen a burgeoning of research attempting to demonstrate that progressive HR practices result in higher organizational performance. Huselid’s (1995)groundbreaking study demonstrated that a set of HR practices he referred to as High Performance Work Systems (HPWS) were related to accounting profits and market value of firms. Since then, a number of studies have shown similar positive relationships between HR practices and various measures of firm performance. While the studies comprising what I refer to as “first generation SHRM research” have added to what is becoming a more convincing body of evidence of the positive relationship between HR and performance, this body tends to lack sufficient data to demonstrate that the relationship is actually causal in the sense that HR practices, when instituted, lead to higher performance. This next generation of SHRM research will begin (and, in fact has begun) to focus on designing more rigorous tests of the hypothesis that employing progressive HRM systems actually results in higher organizational performance. This generation of research will focus on two aspects: demonstrating the HRM value chain, and proving causality as opposed to merely covariation.

Genetic analysis of all-fish growth hormone gene transferred carp (Cyprinus carpio L.) and its F-1 generation

Recombinant "all-fish" growth hormone gene (GH) was microinjected Into the fertilized eggs of carp. A comparison between the growth traits of transgenics and non-transgenics was carried out, and the transgenic individuals with significant "fast-growing" effect were successfully gained. A comparison on the reproductivities was also given out between the transgenics and their non-transgenic siblings, and showed that the reproductive capacity of transgenics was substantially equivalent to those of the non-transgenics. On the other hand, the genetic separation and the characteristic distribution of the F-1 generation were genetically analyzed, which gave solid evidence for the hypothesis that 2-3 chromosomes are integrated with transgene. In addition, the distinct biological effects for multisite-integrated transgenes were further discussed. The present study opens a door for the breeding of "fast-growing" transgenic fish.

Effects of the novel (Pro(3))GIP antagonist and exendin(9-39)amide on GIP- and GLP-1-induced cyclic AMP generation, insulin secretion and postprandial insulin release in obese diabetic (ob/ob) mice: evidence that GIP is the major physiological incretin

AIMS/HYPOTHESIS: This study examined the biological effects of the GIP receptor antagonist, (Pro3)GIP and the GLP-1 receptor antagonist, exendin(9-39)amide.

METHODS: Cyclic AMP production was assessed in Chinese hamster lung fibroblasts transfected with human GIP or GLP-1 receptors, respectively. In vitro insulin release studies were assessed in BRIN-BD11 cells while in vivo insulinotropic and glycaemic responses were measured in obese diabetic ( ob/ ob) mice.

RESULTS: In GIP receptor-transfected fibroblasts, (Pro(3))GIP or exendin(9-39)amide inhibited GIP-stimulated cyclic AMP production with maximal inhibition of 70.0+/-3.5% and 73.5+/-3.2% at 10(-6) mol/l, respectively. In GLP-1 receptor-transfected fibroblasts, exendin(9-39)amide inhibited GLP-1-stimulated cyclic AMP production with maximal inhibition of 60+/-0.7% at 10(-6) mol/l, whereas (Pro(3))GIP had no effect. (Pro(3))GIP specifically inhibited GIP-stimulated insulin release (86%; p<0.001) from clonal BRIN-BD11 cells, but had no effect on GLP-1-stimulated insulin release. In contrast, exendin(9-39)amide inhibited both GIP and GLP-1-stimulated insulin release (57% and 44%, respectively; p<0.001). Administration of (Pro(3))GIP, exendin(9-39)amide or a combination of both peptides (25 nmol/kg body weight, i.p.) to fasted (ob/ob) mice decreased the plasma insulin responses by 42%, 54% and 49%, respectively (p<0.01 to p<0.001). The hyperinsulinaemia of non-fasted (ob/ob) mice was decreased by 19%, 27% and 18% (p<0.05 to p<0.01) by injection of (Pro3)GIP, exendin(9-39)amide or combined peptides but accompanying changes of plasma glucose were small.

CONCLUSIONS/INTERPRETATION: These data show that (Pro(3))GIP is a specific GIP receptor antagonist. Furthermore, feeding studies in one commonly used animal model of obesity and diabetes, (ob/ob) mice, suggest that GIP is the major physiological component of the enteroinsular axis, contributing approximately 80% to incretin-induced insulin release.

Size-assortative pairing in Gammarus pulex (Crustacea : Amphipoda): a test of the timing hypothesis

We present the first empirical test of the timing hypothesis regarding the generation of size-assortative pairing in amphipods. The timing hypothesis proposes that, since large males are better able to afford the costs of mate guarding than small males, the former can take larger females into precopula earlier in the female moult cycle than is feasible for the latter. This leaves small males to form pairs with smaller females closer to moult, thus generating size assortment. We presented male Gammarus pulex, collected both in precopula and as singletons, with females that were (1) previously guarded and therefore near to copulatory moult and (2) previously unguarded and therefore far from copulatory moult. This comparison tested the prediction of the timing hypothesis, that size assortment should break down when the opportunity for time-based male decisions is removed, but that size assortment should occur where timing is not disrupted. Counter to the hypothesis, we found that size assortment did not break down upon removal of the time factor. Large males tended to initiate mate guarding earlier than small males in both female moult groups. However, only in the previously unguarded group did large males guard for longer than small males. This result suggests that, although size assortment occurred in all groups, the causative mechanisms that generated this pattern may differ between these groups. We therefore consider the possible importance of mechanisms such as aggression, simultaneous manipulation of females and female resistance in producing size assortment where males encounter numerous females that are close to moult. We also observed that prior recent guarding experience by males had no effect on latency to guard or size-assortative pairing. (C) 2002 The Association for the Study of Animal Behaviour. Published by Elsevier Science Ltd. All rights reserved.

Advancing impact prediction and hypothesis testing in invasion ecology using a comparative functional response approach

Invasion ecology urgently requires predictive methodologies that can forecast the ecological impacts of existing, emerging and potential invasive species. We argue that many ecologically damaging invaders are characterised by their more efficient use of resources. Consequently, comparison of the classical ‘functional response’ (relationship between resource use and availability) between invasive and trophically analogous native species may allow prediction of invader ecological impact. We review the utility of species trait comparisons and the history and context of the use of functional responses in invasion ecology, then present our framework for the use of comparative functional responses. We show that functional response analyses, by describing the resource use of species over a range of resource availabilities, avoids many pitfalls of ‘snapshot’ assessments of resource use. Our framework demonstrates how comparisons of invader and native functional responses, within and between Type II and III functional responses, allow testing of the likely population-level outcomes of invasions for affected species. Furthermore, we describe how recent studies support the predictive capacity of this method; for example, the invasive ‘bloody red shrimp’ Hemimysis anomala shows higher Type II functional responses than native mysids and this corroborates, and could have predicted, actual invader impacts in the field. The comparative functional response method can also be used to examine differences in the impact of two or more invaders, two or more populations of the same invader, and the abiotic (e.g. temperature) and biotic (e.g. parasitism) context-dependencies of invader impacts. Our framework may also address the previous lack of rigour in testing major hypotheses in invasion ecology, such as the ‘enemy release’ and ‘biotic resistance’ hypotheses, as our approach explicitly considers demographic consequences for impacted resources, such as native and invasive prey species. We also identify potential challenges in the application of comparative functional responses in invasion ecology. These include incorporation of numerical responses, multiple predator effects and trait-mediated indirect interactions, replacement versus non-replacement study designs and the inclusion of functional responses in risk assessment frameworks. In future, the generation of sufficient case studies for a meta-analysis could test the overall hypothesis that comparative functional responses can indeed predict invasive species impacts.

Transcription profiling of Prss16 (Tssp) can be used to find additional peptidase genes that are candidates for self-peptide generation in the thymus

Positive selection (PS) in the thymus involves the presentation of self-peptides that are bound to MHC class II on the surface of cortical thymus epithelial cells (cTECs). Prss16 gene corresponds to one important element regulating the PS of CD4(+) T lymphocytes, which encodes Thymus-specific serine protease (Tssp), a cTEC serine-type peptidase involved in the proteolytic generation of self-peptides. Nevertheless, additional peptidase genes participating in the generation of self-peptides need to be found. Because of its role in the mechanism of PS and its expression in cTECs, the Prss16 gene might be used as a transcriptional marker to identify new genes that share the same expression profile and that encode peptidases in the thymus. To test this hypothesis, we compared the differential thymic expression of 4,500 mRNAs of wild-type (WT) C57BL/6 mice with their respective Prss16-knockout (KO) mutants by using microarrays. From these, 223 genes were differentially expressed, of which 115 had known molecular/biological functions. Four endopeptidase genes (Casp1, Casp2, Psmb3 and Tpp2) share the same expression profile as the Prss16 gene; i.e., induced in WT and repressed in KO while one endopeptidase gene, Capns1, features opposite expression profile. The Tpp2 gene is highlighted because it encodes a serine-type endopeptidase functionally similar to the Tssp enzyme. Profiling of the KO mice featured down-regulation of Prss16, as expected, along with the genes mentioned above. Considering that the Prss16-KO mice featured impaired PS, the shared regulation of the four endopeptidase genes suggested their participation in the mechanism of self-peptide generation and PS.

The lectin-like domain of tumor necrosis factor improves lung function after rat lung transplantation--potential role for a reduction in reactive oxygen species generation

To test the hypothesis that the lectin-like domain of tumor necrosis factor, mimicked by the TIP peptide, can improve lung function after unilateral orthotopic lung isotransplantation. Because of a lack of a specific treatment for ischemia reperfusion-mediated lung injury, accompanied by a disrupted barrier integrity and a dysfunctional alveolar liquid clearance, alternative therapies restoring these parameters after lung transplantation are required.

Efficacy of New Generation Antidepressants: Differences Seem Illusory

Background: Recently, Cipriani and colleagues examined the relative efficacy of 12 new-generation antidepressants on major depression using network meta-analytic methods. They found that some of these medications outperformed others in patient response to treatment. However, several methodological criticisms have been raised about network meta-analysis and Cipriani’s analysis in particular which creates the concern that the stated superiority of some antidepressants relative to others may be unwarranted. Materials and Methods: A Monte Carlo simulation was conducted which involved replicating Cipriani’s network metaanalysis under the null hypothesis (i.e., no true differences between antidepressants). The following simulation strategy was implemented: (1) 1000 simulations were generated under the null hypothesis (i.e., under the assumption that there were no differences among the 12 antidepressants), (2) each of the 1000 simulations were network meta-analyzed, and (3) the total number of false positive results from the network meta-analyses were calculated. Findings: Greater than 7 times out of 10, the network meta-analysis resulted in one or more comparisons that indicated the superiority of at least one antidepressant when no such true differences among them existed. Interpretation: Based on our simulation study, the results indicated that under identical conditions to those of the 117 RCTs with 236 treatment arms contained in Cipriani et al.’s meta-analysis, one or more false claims about the relative efficacy of antidepressants will be made over 70% of the time. As others have shown as well, there is little evidence in these trials that any antidepressant is more effective than another. The tendency of network meta-analyses to generate false positive results should be considered when conducting multiple comparison analyses.

Age-Independent Cartilage Generation for Synovium-Based Autologous Chondrocyte Implantation.

The articular cartilage layer of synovial joints is commonly lesioned by trauma or by a degenerative joint disease. Attempts to repair the damage frequently involve the performance of autologous chondrocyte implantation (ACI). Healthy cartilage must be first removed from the joint, and then, on a separate occasion, following the isolation of the chondrocytes and their expansion in vitro, implanted within the lesion. The disadvantages of this therapeutic approach include the destruction of healthy cartilage-which may predispose the joint to osteoarthritic degeneration-the necessarily restricted availability of healthy tissue, the limited proliferative capacity of the donor cells-which declines with age-and the need for two surgical interventions. We postulated that it should be possible to induce synovial stem cells, which are characterized by high, age-independent, proliferative and chondrogenic differentiation capacities, to lay down cartilage within the outer juxtasynovial space after the transcutaneous implantation of a carrier bearing BMP-2 in a slow-release system. The chondrocytes could be isolated on-site and immediately used for ACI. To test this hypothesis, Chinchilla rabbits were used as an experimental model. A collagenous patch bearing BMP-2 in a slow-delivery vehicle was sutured to the inner face of the synovial membrane. The neoformed tissue was excised 5, 8, 11 and 14 days postimplantation for histological and histomorphometric analyses. Neoformed tissue was observed within the outer juxtasynovial space already on the 5th postimplantation day. It contained connective and adipose tissues, and a central nugget of growing cartilage. Between days 5 and 14, the absolute volume of cartilage increased, attaining a value of 12 mm(3) at the latter juncture. Bone was deposited in measurable quantities from the 11th day onwards, but owing to resorption, the net volume did not exceed 1.5 mm(3) (14th day). The findings confirm our hypothesis. The quantity of neoformed cartilage that is deposited after only 1 week within the outer juxtasynovial space would yield sufficient cells for ACI. Since the BMP-2-bearing patches would be implanted transcutaneously in humans, only one surgical or arthroscopic intervention would be called for. Moreover, most importantly, sufficient numbers of cells could be generated in patients of all ages.

BAC-VAC, a novel generation of (DNA) vaccines: A bacterial artificial chromosome (BAC) containing a replication-competent, packaging-defective virus genome induces protective immunity against herpes simplex virus 1

This study aimed to exploit bacterial artificial chromosomes (BAC) as large antigen-capacity DNA vaccines (BAC-VAC) against complex pathogens, such as herpes simplex virus 1 (HSV-1). The 152-kbp HSV-1 genome recently has been cloned as an F-plasmid-based BAC in Escherichia coli (fHSV), which can efficiently produce infectious virus progeny upon transfection into mammalian cells. A safe modification of fHSV, fHSVΔpac, does not give rise to progeny virus because the signals necessary to package DNA into virions have been excluded. However, in mammalian cells fHSVΔpac DNA can still replicate, express the HSV-1 genes, cause cytotoxic effects, and produce virus-like particles. Because these functions mimic the lytic cycle of the HSV-1 infection, fHSVΔpac was expected to stimulate the immune system as efficiently as a modified live virus vaccine. To test this hypothesis, mice were immunized with fHSVΔpac DNA applied intradermally by gold-particle bombardment, and the immune responses were compared with those induced by infection with disabled infectious single cycle HSV-1. Immunization with either fHSVΔpac or disabled infectious single cycle HSV-1 induced the priming of HSV-1-specific cytotoxic T cells and the production of virus-specific antibodies and conferred protection against intracerebral injection of wild-type HSV-1 at a dose of 200 LD50. Protection probably was cell-mediated, as transfer of serum from immunized mice did not protect naive animals. We conclude that BAC-VACs per se, or in combination with genetic elements that support replicative amplification of the DNA in the cell nucleus, represent a useful new generation of DNA-based vaccination strategies for many viral and nonviral antigens.

Generation of in vivo activating factors in the ischemic intestine by pancreatic enzymes

One of the early events in physiological shock is the generation of activators for leukocytes, endothelial cells, and other cells in the cardiovascular system. The mechanism by which these activators are produced has remained unresolved. We examine here the hypothesis that pancreatic digestive enzymes in the ischemic intestine may be involved in the generation of activators during intestinal ischemia. The lumen of the small intestine of rats was continuously perfused with saline containing a broadly acting pancreatic enzyme inhibitor (6-amidino-2-naphthyl p-guanidinobenzoate dimethanesulfate, 0.37 mM) before and during ischemia of the small intestine by splanchnic artery occlusion. This procedure inhibited activation of circulating leukocytes during occlusion and reperfusion. It also prevented the appearance of activators in portal venous and systemic artery plasma and attenuated initiating symptoms of multiple organ injury in shock. Intestinal tissue produces only low levels of activators in the absence of pancreatic enzymes, whereas in the presence of enzymes, activators are produced in a concentration- and time-dependent fashion. The results indicate that pancreatic digestive enzymes in the ischemic intestine serve as an important source for cell activation and inflammation, as well as multiple organ failure.