Analyzing user interactions with biomedical ontologies: A visual perspective.

Biomedical ontologies are large: Several ontologies in the BioPortal repository contain thousands or even hundreds of thousands of entities. The development and maintenance of such large ontologies is difficult. To support ontology authors and repository developers in their work, it is crucial to improve our understanding of how these ontologies are explored, queried, reused, and used in downstream applications by biomedical researchers. We present an exploratory empirical analysis of user activities in the BioPortal ontology repository by analyzing BioPortal interaction logs across different access modes over several years. We investigate how users of BioPortal query and search for ontologies and their classes, how they explore the ontologies, and how they reuse classes from different ontologies. Additionally, through three real-world scenarios, we not only analyze the usage of ontologies for annotation tasks but also compare it to the browsing and querying behaviors of BioPortal users. For our investigation, we use several different visualization techniques. To inspect large amounts of interaction, reuse, and real-world usage data at a glance, we make use of and extend PolygOnto, a visualization method that has been successfully used to analyze reuse of ontologies in previous work. Our results show that exploration, query, reuse, and actual usage behaviors rarely align, suggesting that different users tend to explore, query and use different parts of an ontology. Finally, we highlight and discuss differences and commonalities among users of BioPortal.

The Reactome pathway knowledgebase.

Reactome (http://www.reactome.org) is a manually curated open-source open-data resource of human pathways and reactions. The current version 46 describes 7088 human proteins (34% of the predicted human proteome), participating in 6744 reactions based on data extracted from 15 107 research publications with PubMed links. The Reactome Web site and analysis tool set have been completely redesigned to increase speed, flexibility and user friendliness. The data model has been extended to support annotation of disease processes due to infectious agents and to mutation.

ReVeaLD: a user-driven domain-specific interactive search platform for biomedical research.

Bioinformatics research relies heavily on the ability to discover and correlate data from various sources. The specialization of life sciences over the past decade, coupled with an increasing number of biomedical datasets available through standardized interfaces, has created opportunities towards new methods in biomedical discovery. Despite the popularity of semantic web technologies in tackling the integrative bioinformatics challenge, there are many obstacles towards its usage by non-technical research audiences. In particular, the ability to fully exploit integrated information needs using improved interactive methods intuitive to the biomedical experts. In this report we present ReVeaLD (a Real-time Visual Explorer and Aggregator of Linked Data), a user-centered visual analytics platform devised to increase intuitive interaction with data from distributed sources. ReVeaLD facilitates query formulation using a domain-specific language (DSL) identified by biomedical experts and mapped to a self-updated catalogue of elements from external sources. ReVeaLD was implemented in a cancer research setting; queries included retrieving data from in silico experiments, protein modeling and gene expression. ReVeaLD was developed using Scalable Vector Graphics and JavaScript and a demo with explanatory video is available at http://www.srvgal78.deri.ie:8080/explorer. A set of user-defined graphic rules controls the display of information through media-rich user interfaces. Evaluation of ReVeaLD was carried out as a game: biomedical researchers were asked to assemble a set of 5 challenge questions and time and interactions with the platform were recorded. Preliminary results indicate that complex queries could be formulated under less than two minutes by unskilled researchers. The results also indicate that supporting the identification of the elements of a DSL significantly increased intuitiveness of the platform and usability of semantic web technologies by domain users.

An empirical meta-analysis of the life sciences linked open data on the web.

While the biomedical community has published several "open data" sources in the last decade, most researchers still endure severe logistical and technical challenges to discover, query, and integrate heterogeneous data and knowledge from multiple sources. To tackle these challenges, the community has experimented with Semantic Web and linked data technologies to create the Life Sciences Linked Open Data (LSLOD) cloud. In this paper, we extract schemas from more than 80 biomedical linked open data sources into an LSLOD schema graph and conduct an empirical meta-analysis to evaluate the extent of semantic heterogeneity across the LSLOD cloud. We observe that several LSLOD sources exist as stand-alone data sources that are not inter-linked with other sources, use unpublished schemas with minimal reuse or mappings, and have elements that are not useful for data integration from a biomedical perspective. We envision that the LSLOD schema graph and the findings from this research will aid researchers who wish to query and integrate data and knowledge from multiple biomedical sources simultaneously on the Web.

Using ethnographic methods to classify the human experience in medicine: a case study of the presence ontology.

OBJECTIVE: Although social and environmental factors are central to provider-patient interactions, the data that reflect these factors can be incomplete, vague, and subjective. We sought to create a conceptual framework to describe and classify data about presence, the domain of interpersonal connection in medicine. METHODS: Our top-down approach for ontology development based on the concept of "relationality" included the following: 1) a broad survey of the social sciences literature and a systematic literature review of >20 000 articles around interpersonal connection in medicine, 2) relational ethnography of clinical encounters (n = 5 pilot, 27 full), and 3) interviews about relational work with 40 medical and nonmedical professionals. We formalized the model using the Web Ontology Language in the Protégé ontology editor. We iteratively evaluated and refined the Presence Ontology through manual expert review and automated annotation of literature. RESULTS AND DISCUSSION: The Presence Ontology facilitates the naming and classification of concepts that would otherwise be vague. Our model categorizes contributors to healthcare encounters and factors such as communication, emotions, tools, and environment. Ontology evaluation indicated that cognitive models (both patients' explanatory models and providers' caregiving approaches) influenced encounters and were subsequently incorporated. We show how ethnographic methods based in relationality can aid the representation of experiential concepts (eg, empathy, trust). Our ontology could support investigative methods to improve healthcare processes for both patients and healthcare providers, including annotation of videotaped encounters, development of clinical instruments to measure presence, or implementation of electronic health record-based reminders for providers. CONCLUSION: The Presence Ontology provides a model for using ethnographic approaches to classify interpersonal data.

Text Snippets to Corroborate Medical Relations: An Unsupervised Approach using a Knowledge Graph and Embeddings.

Knowledge graphs have been shown to significantly improve search results. Usually populated by subject matter experts, relations therein need to keep up to date with medical literature in order for search to remain relevant. Dynamically identifying text snippets in literature that confirm or deny knowledge graph triples is increasingly becoming the differentiator between trusted and untrusted medical decision support systems. This work describes our approach to mapping triples to medical text. A medical knowledge graph is used as a source of triples that are used to find matching sentences in reference text. Our unsupervised approach uses phrase embeddings and cosine similarity measures, and boosts candidate text snippets when certain key concepts exist. Using this approach, we can accurately map semantic relations within the medical knowledge graph to text snippets with a precision of 61.4% and recall of 86.3%. This method will be used to develop a novel application in the future to retrieve medical relations and corroborating snippets from medical text given a user query.

Enabling Web-scale data integration in biomedicine through Linked Open Data.

The biomedical data landscape is fragmented with several isolated, heterogeneous data and knowledge sources, which use varying formats, syntaxes, schemas, and entity notations, existing on the Web. Biomedical researchers face severe logistical and technical challenges to query, integrate, analyze, and visualize data from multiple diverse sources in the context of available biomedical knowledge. Semantic Web technologies and Linked Data principles may aid toward Web-scale semantic processing and data integration in biomedicine. The biomedical research community has been one of the earliest adopters of these technologies and principles to publish data and knowledge on the Web as linked graphs and ontologies, hence creating the Life Sciences Linked Open Data (LSLOD) cloud. In this paper, we provide our perspective on some opportunities proffered by the use of LSLOD to integrate biomedical data and knowledge in three domains: (1) pharmacology, (2) cancer research, and (3) infectious diseases. We will discuss some of the major challenges that hinder the wide-spread use and consumption of LSLOD by the biomedical research community. Finally, we provide a few technical solutions and insights that can address these challenges. Eventually, LSLOD can enable the development of scalable, intelligent infrastructures that support artificial intelligence methods for augmenting human intelligence to achieve better clinical outcomes for patients, to enhance the quality of biomedical research, and to improve our understanding of living systems.

MRI to MGMT: predicting methylation status in glioblastoma patients using convolutional recurrent neural networks.

Glioblastoma Multiforme (GBM), a malignant brain tumor, is among the most lethal of all cancers. Temozolomide is the primary chemotherapy treatment for patients diagnosed with GBM. The methylation status of the promoter or the enhancer regions of the O6-methylguanine methyltransferase (MGMT) gene may impact the efficacy and sensitivity of temozolomide, and hence may affect overall patient survival. Microscopic genetic changes may manifest as macroscopic morphological changes in the brain tumors that can be detected using magnetic resonance imaging (MRI), which can serve as noninvasive biomarkers for determining methylation of MGMT regulatory regions. In this research, we use a compendium of brain MRI scans of GBM patients collected from The Cancer Imaging Archive (TCIA) combined with methylation data from The Cancer Genome Atlas (TCGA) to predict the methylation state of the MGMT regulatory regions in these patients. Our approach relies on a bi-directional convolutional recurrent neural network architecture (CRNN) that leverages the spatial aspects of these 3-dimensional MRI scans. Our CRNN obtains an accuracy of 67% on the validation data and 62% on the test data, with precision and recall both at 67%, suggesting the existence of MRI features that may complement existing markers for GBM patient stratification and prognosis. We have additionally presented our model via a novel neural network visualization platform, which we have developed to improve interpretability of deep learning MRI-based classification models.

Mechanism-based Pharmacovigilance over the Life Sciences Linked Open Data Cloud.

Adverse drug reactions (ADR) result in significant morbidity and mortality in patients, and a substantial proportion of these ADRs are caused by drug-drug interactions (DDIs). Pharmacovigilance methods are used to detect unanticipated DDIs and ADRs by mining Spontaneous Reporting Systems, such as the US FDA Adverse Event Reporting System (FAERS). However, these methods do not provide mechanistic explanations for the discovered drug-ADR associations in a systematic manner. In this paper, we present a systems pharmacology-based approach to perform mechanism-based pharmacovigilance. We integrate data and knowledge from four different sources using Semantic Web Technologies and Linked Data principles to generate a systems network. We present a network-based Apriori algorithm for association mining in FAERS reports. We evaluate our method against existing pharmacovigilance methods for three different validation sets. Our method has AUROC statistics of 0.7-0.8, similar to current methods, and event-specific thresholds generate AUROC statistics greater than 0.75 for certain ADRs. Finally, we discuss the benefits of using Semantic Web technologies to attain the objectives for mechanism-based pharmacovigilance.

PhLeGrA: Graph Analytics in Pharmacology over the Web of Life Sciences Linked Open Data.

Integrated approaches for pharmacology are required for the mechanism-based predictions of adverse drug reactions that manifest due to concomitant intake of multiple drugs. These approaches require the integration and analysis of biomedical data and knowledge from multiple, heterogeneous sources with varying schemas, entity notations, and formats. To tackle these integrative challenges, the Semantic Web community has published and linked several datasets in the Life Sciences Linked Open Data (LSLOD) cloud using established W3C standards. We present the PhLeGrA platform for Linked Graph Analytics in Pharmacology in this paper. Through query federation, we integrate four sources from the LSLOD cloud and extract a drug-reaction network, composed of distinct entities. We represent this graph as a hidden conditional random field (HCRF), a discriminative latent variable model that is used for structured output predictions. We calculate the underlying probability distributions in the drug-reaction HCRF using the datasets from the U.S. Food and Drug Administration's Adverse Event Reporting System. We predict the occurrence of 146 adverse reactions due to multiple drug intake with an AUROC statistic greater than 0.75. The PhLeGrA platform can be extended to incorporate other sources published using Semantic Web technologies, as well as to discover other types of pharmacological associations.

A Systematic Analysis of Term Reuse and Term Overlap across Biomedical Ontologies.

Reusing ontologies and their terms is a principle and best practice that most ontology development methodologies strongly encourage. Reuse comes with the promise to support the semantic interoperability and to reduce engineering costs. In this paper, we present a descriptive study of the current extent of term reuse and overlap among biomedical ontologies. We use the corpus of biomedical ontologies stored in the BioPortal repository, and analyze different types of reuse and overlap constructs. While we find an approximate term overlap between 25-31%, the term reuse is only <9%, with most ontologies reusing fewer than 5% of their terms from a small set of popular ontologies. Clustering analysis shows that the terms reused by a common set of ontologies have >90% semantic similarity, hinting that ontology developers tend to reuse terms that are sibling or parent-child nodes. We validate this finding by analysing the logs generated from a Protégé plugin that enables developers to reuse terms from BioPortal. We find most reuse constructs were 2-level subtrees on the higher levels of the class hierarchy. We developed a Web application that visualizes reuse dependencies and overlap among ontologies, and that proposes similar terms from BioPortal for a term of interest. We also identified a set of error patterns that indicate that ontology developers did intend to reuse terms from other ontologies, but that they were using different and sometimes incorrect representations. Our results stipulate the need for semi-automated tools that augment term reuse in the ontology engineering process through personalized recommendations.

BiOnIC: A Catalog of User Interactions with Biomedical Ontologies.

BiOnIC is a catalog of aggregated statistics of user clicks, queries, and reuse counts for access to over 200 biomedical ontologies. BiOnIC also provides anonymized sequences of classes accessed by users over a period of four years. To generate the statistics, we processed the access logs of BioPortal, a large open biomedical ontology repository. We publish the BiOnIC data using DCAT and SKOS metadata standards. The BiOnIC catalog has a wide range of applicability, which we demonstrate through its use in three different types of applications. To our knowledge, this type of interaction data stemming from a real-world, large-scale application has not been published before. We expect that the catalog will become an important resource for researchers and developers in the Semantic Web community by providing novel insights into how ontologies are explored, queried and reused. The BiOnIC catalog may ultimately assist in the more informed development of intelligent user interfaces for semantic resources through interface customization, prediction of user browsing and querying behavior, and ontology summarization. The BiOnIC catalog is available at: http://onto-apps.stanford.edu/bionic.

PRISM: A DATA-DRIVEN PLATFORM FOR MONITORING MENTAL HEALTH.

Neuropsychiatric disorders are the leading cause of disability worldwide and there is no gold standard currently available for the measurement of mental health. This issue is exacerbated by the fact that the information physicians use to diagnose these disorders is episodic and often subjective. Current methods to monitor mental health involve the use of subjective DSM-5 guidelines, and advances in EEG and video monitoring technologies have not been widely adopted due to invasiveness and inconvenience. Wearable technologies have surfaced as a ubiquitous and unobtrusive method for providing continuous, quantitative data about a patient. Here, we introduce PRISM-Passive, Real-time Information for Sensing Mental Health. This platform integrates motion, light and heart rate data from a smart watch application with user interactions and text entries from a web application. We have demonstrated a proof of concept by collecting preliminary data through a pilot study of 13 subjects. We have engineered appropriate features and applied both unsupervised and supervised learning to develop models that are predictive of user-reported ratings of their emotional state, demonstrating that the data has the potential to be useful for evaluating mental health. This platform could allow patients and clinicians to leverage continuous streams of passive data for early and accurate diagnosis as well as constant monitoring of patients suffering from mental disorders.

An Ebola virus-centered knowledge base.

Ebola virus (EBOV), of the family Filoviridae viruses, is a NIAID category A, lethal human pathogen. It is responsible for causing Ebola virus disease (EVD) that is a severe hemorrhagic fever and has a cumulative death rate of 41% in the ongoing epidemic in West Africa. There is an ever-increasing need to consolidate and make available all the knowledge that we possess on EBOV, even if it is conflicting or incomplete. This would enable biomedical researchers to understand the molecular mechanisms underlying this disease and help develop tools for efficient diagnosis and effective treatment. In this article, we present our approach for the development of an Ebola virus-centered Knowledge Base (Ebola-KB) using Linked Data and Semantic Web Technologies. We retrieve and aggregate knowledge from several open data sources, web services and biomedical ontologies. This knowledge is transformed to RDF, linked to the Bio2RDF datasets and made available through a SPARQL 1.1 Endpoint. Ebola-KB can also be explored using an interactive Dashboard visualizing the different perspectives of this integrated knowledge. We showcase how different competency questions, asked by domain users researching the druggability of EBOV, can be formulated as SPARQL Queries or answered using the Ebola-KB Dashboard.

Investigating Term Reuse and Overlap in Biomedical Ontologies.

We investigate the current extent of term reuse and overlap among biomedical ontologies. We use the corpus of biomedical ontologies stored in the BioPortal repository, and analyze three types of reuse constructs: (a) explicit term reuse, (b) xref reuse, and (c) Concept Unique Identifier (CUI) reuse. While there is a term label similarity of approximately 14.4% of the total terms, we observed that most ontologies reuse considerably fewer than 5% of their terms from a concise set of a few core ontologies. We developed an interactive visualization to explore reuse dependencies among biomedical ontologies. Moreover, we identified a set of patterns that indicate ontology developers did intend to reuse terms from other ontologies, but they were using different and sometimes incorrect representations. Our results suggest the value of semi-automated tools that augment term reuse in the ontology engineering process through personalized recommendations.

Functional characterization of two structurally novel diacylglycerol acyltransferase2 isozymes responsible for the enhanced production of stearate-rich storage lipid in Candida tropicalis SY005.

Diacylglycerol acyltransferase (DGAT) activity is an essential enzymatic step in the formation of neutral lipid i.e., triacylglycerol in all living cells capable of accumulating storage lipid. Previously, we characterized an oleaginous yeast Candida tropicalis SY005 that yields storage lipid up to 58% under a specific nitrogen-stress condition, when the DGAT-specific transcript is drastically up-regulated. Here we report the identification, differential expression and function of two DGAT2 gene homologues--CtDGAT2a and CtDGAT2b of this C. tropicalis. Two protein isoforms are unique with respect to the presence of five additional stretches of amino acids, besides possessing three highly conserved motifs known in other reported DGAT2 enzymes. Moreover, the CtDGAT2a and CtDGAT2b are characteristically different in amino acid sequences and predicted protein structures. The CtDGAT2b isozyme was found to be catalytically 12.5% more efficient than CtDGAT2a for triacylglycerol production in a heterologous yeast system i.e., Saccharomyces cerevisiae quadruple mutant strain H1246 that is inherently defective in neutral lipid biosynthesis. The CtDGAT2b activity rescued the growth of transformed S. cerevisiae mutant cells, which are usually non-viable in the medium containing free fatty acids by incorporating them into triacylglycerol, and displayed preferential specificity towards saturated acyl species as substrate. Furthermore, we document that the efficiency of triacylglycerol production by CtDGAT2b is differentially affected by deletion, insertion or replacement of amino acids in five regions exclusively present in two CtDGAT2 isozymes. Taken together, our study characterizes two structurally novel DGAT2 isozymes, which are accountable for the enhanced production of storage lipid enriched with saturated fatty acids inherently in C. tropicalis SY005 strain as well as in transformed S. cerevisiae neutral lipid-deficient mutant cells. These two genes certainly will be useful for further investigation on the novel structure-function relationship of DGAT repertoire, and also in metabolic engineering for the enhanced production of lipid feedstock in other organisms.

Identification of an extracellular antifungal protein from the endophytic fungus Colletotrichum sp. DM06.

An extracellular antifungal protein of 28 kDa (exAFP-C28) was identified from an endophytic fungus Colletotrichum sp. DM-06. After purification, the MIC value of exAFP-C28 against Candida albicans, a well-known human pathogenic fungus was found to be 32 µg/mL that unaffected the human red blood cells. The antifungal activity associated with exAFP-C28 was manifested by the increased membrane permeability of C. albicans cells followed by disruption. Proteomics and bioinformatics analyses revealed that several peptide fragments of exAFP-C28 have identity with the bacterial 50S ribosomal protein L10, and a stretch of 55 amino acids of two peptide fragments corresponding to the Nterminus of L10 protein is capable of forming amphipathic helix required for membrane penetration. Taken together, our results suggest that the exAFP-C28 protein from Colletotrichum sp. DM-06 is a promising therapeutic agent in controlling candidiasis disease in animals including humans.