Pressing needs of biomedical text mining in biocuration and beyond: opportunities and challenges.

Text mining in the biomedical sciences is rapidly transitioning from small-scale evaluation to large-scale application. In this article, we argue that text-mining technologies have become essential tools in real-world biomedical research. We describe four large scale applications of text mining, as showcased during a recent panel discussion at the BioCreative V Challenge Workshop. We draw on these applications as case studies to characterize common requirements for successfully applying text-mining techniques to practical biocuration needs. We note that system 'accuracy' remains a challenge and identify several additional common difficulties and potential research directions including (i) the 'scalability' issue due to the increasing need of mining information from millions of full-text articles, (ii) the 'interoperability' issue of integrating various text-mining systems into existing curation workflows and (iii) the 'reusability' issue on the difficulty of applying trained systems to text genres that are not seen previously during development. We then describe related efforts within the text-mining community, with a special focus on the BioCreative series of challenge workshops. We believe that focusing on the near-term challenges identified in this work will amplify the opportunities afforded by the continued adoption of text-mining tools. Finally, in order to sustain the curation ecosystem and have text-mining systems adopted for practical benefits, we call for increased collaboration between text-mining researchers and various stakeholders, including researchers, publishers and biocurators.

Training and evaluation corpora for the extraction of causal relationships encoded in biological expression language (BEL).

Success in extracting biological relationships is mainly dependent on the complexity of the task as well as the availability of high-quality training data. Here, we describe the new corpora in the systems biology modeling language BEL for training and testing biological relationship extraction systems that we prepared for the BioCreative V BEL track. BEL was designed to capture relationships not only between proteins or chemicals, but also complex events such as biological processes or disease states. A BEL nanopub is the smallest unit of information and represents a biological relationship with its provenance. In BEL relationships (called BEL statements), the entities are normalized to defined namespaces mainly derived from public repositories, such as sequence databases, MeSH or publicly available ontologies. In the BEL nanopubs, the BEL statements are associated with citation information and supportive evidence such as a text excerpt. To enable the training of extraction tools, we prepared BEL resources and made them available to the community. We selected a subset of these resources focusing on a reduced set of namespaces, namely, human and mouse genes, ChEBI chemicals, MeSH diseases and GO biological processes, as well as relationship types 'increases' and 'decreases'. The published training corpus contains 11 000 BEL statements from over 6000 supportive text excerpts. For method evaluation, we selected and re-annotated two smaller subcorpora containing 100 text excerpts. For this re-annotation, the inter-annotator agreement was measured by the BEL track evaluation environment and resulted in a maximal F-score of 91.18% for full statement agreement. In addition, for a set of 100 BEL statements, we do not only provide the gold standard expert annotations, but also text excerpts pre-selected by two automated systems. Those text excerpts were evaluated and manually annotated as true or false supportive in the course of the BioCreative V BEL track task.Database URL: http://wiki.openbel.org/display/BIOC/Datasets.

Reverse causal reasoning: applying qualitative causal knowledge to the interpretation of high-throughput data.

BACKGROUND: Gene expression profiling and other genome-scale measurement technologies provide comprehensive information about molecular changes resulting from a chemical or genetic perturbation, or disease state. A critical challenge is the development of methods to interpret these large-scale data sets to identify specific biological mechanisms that can provide experimentally verifiable hypotheses and lead to the understanding of disease and drug action. RESULTS: We present a detailed description of Reverse Causal Reasoning (RCR), a reverse engineering methodology to infer mechanistic hypotheses from molecular profiling data. This methodology requires prior knowledge in the form of small networks that causally link a key upstream controller node representing a biological mechanism to downstream measurable quantities. These small directed networks are generated from a knowledge base of literature-curated qualitative biological cause-and-effect relationships expressed as a network. The small mechanism networks are evaluated as hypotheses to explain observed differential measurements. We provide a simple implementation of this methodology, Whistle, specifically geared towards the analysis of gene expression data and using prior knowledge expressed in Biological Expression Language (BEL). We present the Whistle analyses for three transcriptomic data sets using a publically available knowledge base. The mechanisms inferred by Whistle are consistent with the expected biology for each data set. CONCLUSIONS: Reverse Causal Reasoning yields mechanistic insights to the interpretation of gene expression profiling data that are distinct from and complementary to the results of analyses using ontology or pathway gene sets. This reverse engineering algorithm provides an evidence-driven approach to the development of models of disease, drug action, and drug toxicity.

Comparative transcriptional network modeling of three PPAR-α/γ co-agonists reveals distinct metabolic gene signatures in primary human hepatocytes.

AIMS: To compare the molecular and biologic signatures of a balanced dual peroxisome proliferator-activated receptor (PPAR)-α/γ agonist, aleglitazar, with tesaglitazar (a dual PPAR-α/γ agonist) or a combination of pioglitazone (Pio; PPAR-γ agonist) and fenofibrate (Feno; PPAR-α agonist) in human hepatocytes. METHODS AND RESULTS: Gene expression microarray profiles were obtained from primary human hepatocytes treated with EC(50)-aligned low, medium and high concentrations of the three treatments. A systems biology approach, Causal Network Modeling, was used to model the data to infer upstream molecular mechanisms that may explain the observed changes in gene expression. Aleglitazar, tesaglitazar and Pio/Feno each induced unique transcriptional signatures, despite comparable core PPAR signaling. Although all treatments inferred qualitatively similar PPAR-α signaling, aleglitazar was inferred to have greater effects on high- and low-density lipoprotein cholesterol levels than tesaglitazar and Pio/Feno, due to a greater number of gene expression changes in pathways related to high-density and low-density lipoprotein metabolism. Distinct transcriptional and biologic signatures were also inferred for stress responses, which appeared to be less affected by aleglitazar than the comparators. In particular, Pio/Feno was inferred to increase NFE2L2 activity, a key component of the stress response pathway, while aleglitazar had no significant effect. All treatments were inferred to decrease proliferative signaling. CONCLUSIONS: Aleglitazar induces transcriptional signatures related to lipid parameters and stress responses that are unique from other dual PPAR-α/γ treatments. This may underlie observed favorable changes in lipid profiles in animal and clinical studies with aleglitazar and suggests a differentiated gene profile compared with other dual PPAR-α/γ agonist treatments.

Early patient stratification and predictive biomarkers in drug discovery and development: a case study of ulcerative colitis anti-TNF therapy.

The current drug discovery paradigm is long, costly, and prone to failure. For projects in early development, lack of efficacy in Phase II is a major contributor to the overall failure rate. Efficacy failures often occur from one of two major reasons: either the investigational agent did not achieve the required pharmacology or the mechanism targeted by the investigational agent did not significantly contribute to the disease in the tested patient population. The latter scenario can arise due to insufficient study power stemming from patient heterogeneity. If the subset of disease patients driven by the mechanism that is likely to respond to the drug can be identified and selected before enrollment begins, efficacy and response rates should improve. This will not only augment drug approval percentages, but will also minimize the number of patients at risk of side effects in the face of a suboptimal response to treatment. Here we describe a systems biology approach using molecular profiling data from patients at baseline for the development of predictive biomarker content to identify potential responders to a molecular targeted therapy before the drug is tested in humans. A case study is presented where a classifier to predict response to a TNF targeted therapy for ulcerative colitis is developed a priori and verified against a test set of patients where clinical outcomes are known. This approach will promote the tandem development of drugs with predictive response, patient selection biomarkers.

A computable cellular stress network model for non-diseased pulmonary and cardiovascular tissue.

BACKGROUND: Humans and other organisms are equipped with a set of responses that can prevent damage from exposure to a multitude of endogenous and environmental stressors. If these stress responses are overwhelmed, this can result in pathogenesis of diseases, which is reflected by an increased development of, e.g., pulmonary and cardiac diseases in humans exposed to chronic levels of environmental stress, including inhaled cigarette smoke (CS). Systems biology data sets (e.g., transcriptomics, phosphoproteomics, metabolomics) could enable comprehensive investigation of the biological impact of these stressors. However, detailed mechanistic networks are needed to determine which specific pathways are activated in response to different stressors and to drive the qualitative and eventually quantitative assessment of these data. A current limiting step in this process is the availability of detailed mechanistic networks that can be used as an analytical substrate. RESULTS: We have built a detailed network model that captures the biology underlying the physiological cellular response to endogenous and exogenous stressors in non-diseased mammalian pulmonary and cardiovascular cells. The contents of the network model reflect several diverse areas of signaling, including oxidative stress, hypoxia, shear stress, endoplasmic reticulum stress, and xenobiotic stress, that are elicited in response to common pulmonary and cardiovascular stressors. We then tested the ability of the network model to identify the mechanisms that are activated in response to CS, a broad inducer of cellular stress. Using transcriptomic data from the lungs of mice exposed to CS, the network model identified a robust increase in the oxidative stress response, largely mediated by the anti-oxidant NRF2 pathways, consistent with previous reports on the impact of CS exposure in the mammalian lung. CONCLUSIONS: The results presented here describe the construction of a cellular stress network model and its application towards the analysis of environmental stress using transcriptomic data. The proof-of-principle analysis described here, coupled with the future development of additional network models covering distinct areas of biology, will help to further clarify the integrated biological responses elicited by complex environmental stressors such as CS, in pulmonary and cardiovascular cells.

Construction of a computable cell proliferation network focused on non-diseased lung cells.

BACKGROUND: Critical to advancing the systems-level evaluation of complex biological processes is the development of comprehensive networks and computational methods to apply to the analysis of systems biology data (transcriptomics, proteomics/phosphoproteomics, metabolomics, etc.). Ideally, these networks will be specifically designed to capture the normal, non-diseased biology of the tissue or cell types under investigation, and can be used with experimentally generated systems biology data to assess the biological impact of perturbations like xenobiotics and other cellular stresses. Lung cell proliferation is a key biological process to capture in such a network model, given the pivotal role that proliferation plays in lung diseases including cancer, chronic obstructive pulmonary disease (COPD), and fibrosis. Unfortunately, no such network has been available prior to this work. RESULTS: To further a systems-level assessment of the biological impact of perturbations on non-diseased mammalian lung cells, we constructed a lung-focused network for cell proliferation. The network encompasses diverse biological areas that lead to the regulation of normal lung cell proliferation (Cell Cycle, Growth Factors, Cell Interaction, Intra- and Extracellular Signaling, and Epigenetics), and contains a total of 848 nodes (biological entities) and 1597 edges (relationships between biological entities). The network was verified using four published gene expression profiling data sets associated with measured cell proliferation endpoints in lung and lung-related cell types. Predicted changes in the activity of core machinery involved in cell cycle regulation (RB1, CDKN1A, and MYC/MYCN) are statistically supported across multiple data sets, underscoring the general applicability of this approach for a network-wide biological impact assessment using systems biology data. CONCLUSIONS: To the best of our knowledge, this lung-focused Cell Proliferation Network provides the most comprehensive connectivity map in existence of the molecular mechanisms regulating cell proliferation in the lung. The network is based on fully referenced causal relationships obtained from extensive evaluation of the literature. The computable structure of the network enables its application to the qualitative and quantitative evaluation of cell proliferation using systems biology data sets. The network is available for public use.

A point mutation in the cargo-binding domain of myosin V affects its interaction with multiple cargoes.

Class V myosins move diverse intracellular cargoes, which attach via interaction of cargo-specific proteins to the myosin V globular tail. The globular tail of the yeast myosin V, Myo2p, contains two structural and functional subdomains. Subdomain I binds to the vacuole-specific protein, Vac17p, while subdomain II likely binds to an as yet unidentified secretory vesicle-specific protein. All functions of Myo2p require the tight association of subdomains I and II, which suggests that binding of a cargo to one subdomain may inhibit cargo-binding to a second subdomain. Thus, two types of mutations are predicted to specifically affect a subset of Myo2p cargoes: first are mutations within a cargo-specific binding region; second are mutations that mimic the inhibited conformation of one of the subdomains. Here we analyze a point mutation in subdomain I, myo2-2(G1248D), which is likely to be this latter type of mutation. myo2-2 has no effect on secretory vesicle movement. The secretory vesicle binding site is in subdomain II. However, myo2-2 is impaired in several Myo2p-related functions. While subdomains I and II of myo2-2p tightly associate, there are measurable differences in the conformation of its globular tail. Based solely on the ability to restore vacuole inheritance, a set of intragenic suppressors of myo2-2 were identified. All suppressor mutations reside in subdomain I. Moreover, subdomain I and II interactions occurred in all suppressors, demonstrating the importance of subdomain I and II association for Myo2p function. Furthermore, 3 of the 10 suppressors globally restored all tested defects in myo2-2. This large proportion of global suppressors strongly suggests that myo2-2(G1248) causes a conformational change in subdomain I that simultaneously affects multiple cargoes.

Myosin V attachment to cargo requires the tight association of two functional subdomains.

The myosin V carboxyl-terminal globular tail domain is essential for the attachment of myosin V to all known cargoes. Previously, the globular tail was viewed as a single, functional entity. Here, we show that the globular tail of the yeast myosin Va homologue, Myo2p, contains two structural subdomains that have distinct functions, namely, vacuole-specific and secretory vesicle-specific movement. Biochemical and genetic analyses demonstrate that subdomain I tightly associates with subdomain II, and that the interaction does not require additional proteins. Importantly, although neither subdomain alone is functional, simultaneous expression of the separate subdomains produces a functional complex in vivo. Our results suggest a model whereby intramolecular interactions between the globular tail subdomains help to coordinate the transport of multiple distinct cargoes by myosin V.

Whole-genome analysis of two-component signal transduction genes in fungal pathogens.

Two-component phosphorelay systems are minimally comprised of a histidine kinase (HK) component, which autophosphorylates in response to an environmental stimulus, and a response regulator (RR) component, which transmits the signal, resulting in an output such as activation of transcription, or of a mitogen-activated protein kinase cascade. The genomes of the yeasts Saccharomyces cerevisiae, Schizosaccharomyces pombe, and Candida albicans encode one, three, and three HKs, respectively. In contrast, the genome sequences of the filamentous ascomycetes Neurospora crassa, Cochliobolus heterostrophus (Bipolaris maydis), Gibberella moniliformis (Fusarium verticillioides), and Botryotinia fuckeliana (Botrytis cinerea) encode an extensive family of two-component signaling proteins. The putative HKs fall into 11 classes. Most of these classes are represented in each filamentous ascomycete species examined. A few of these classes are significantly more prevalent in the fungal pathogens than in the saprobe N. crassa, suggesting that these groups contain paralogs required for virulence. Despite the larger numbers of HKs in filamentous ascomycetes than in yeasts, all of the ascomycetes contain virtually the same downstream histidine phosphotransfer proteins and RR proteins, suggesting extensive cross talk or redundancy among HKs.

Identification of an organelle-specific myosin V receptor.

Class V myosins are widely distributed among diverse organisms and move cargo along actin filaments. Some myosin Vs move multiple types of cargo, where the timing of movement and the destinations of selected cargoes are unique. Here, we report the discovery of an organelle-specific myosin V receptor. Vac17p, a novel protein, is a component of the vacuole-specific receptor for Myo2p, a Saccharomyces cerevisiae myosin V. Vac17p interacts with the Myo2p cargo-binding domain, but not with vacuole inheritance-defective myo2 mutants that have single amino acid changes within this region. Moreover, a region of the Myo2p tail required specifically for secretory vesicle transport is neither required for vacuole inheritance nor for Vac17p-Myo2p interactions. Vac17p is localized on the vacuole membrane, and vacuole-associated Myo2p increases in proportion with an increase in Vac17p. Furthermore, Vac17p is not required for movement of other cargo moved by Myo2p. These findings demonstrate that Vac17p is a component of a vacuole-specific receptor for Myo2p. Organelle-specific receptors such as Vac17p provide a mechanism whereby a single type of myosin V can move diverse cargoes to distinct destinations at different times.

Regulated degradation of a class V myosin receptor directs movement of the yeast vacuole.

Normal cellular function requires that organelles be positioned in specific locations. The direction in which molecular motors move organelles is based in part on the polarity of microtubules and actin filaments. However, this alone does not determine the intracellular destination of organelles. For example, the yeast class V myosin, Myo2p, moves several organelles to distinct locations during the cell cycle. Thus the movement of each type of Myo2p cargo must be regulated uniquely. Here we report a regulatory mechanism that specifically provides directionality to vacuole movement. The vacuole-specific Myo2p receptor, Vac17p, has a key function in this process. Vac17p binds simultaneously to Myo2p and to Vac8p, a vacuolar membrane protein. The transport complex, Myo2p-Vac17p-Vac8p, moves the vacuole to the bud, and is then disrupted through the degradation of Vac17p. The vacuole is ultimately deposited near the centre of the bud. Removal of a PEST sequence (a potential signal for rapid protein degradation) within Vac17p causes its stabilization and the subsequent 'backward' movement of vacuoles, which mis-targets them to the neck between the mother cell and the bud. Thus the regulated disruption of this transport complex places the vacuole in its proper location. This may be a general mechanism whereby organelles are deposited at their terminal destination.