Dynamic enhancement of drug product labels to support drug safety, efficacy, and effectiveness

BETA: a comprehensive benchmark for computational drug-target prediction

Internal validation is the most popular evaluation strategy used for drug-target predictive models. The simple random shuffling in the cross-validation, however, is not always ideal to handle large, diverse and copious datasets as it could potentially introduce bias. Hence, these predictive models cannot be comprehensively evaluated to provide insight into their general performance on a variety of use-cases (e.g. permutations of different levels of connectiveness and categories in drug and target space, as well as validations based on different data sources). In this work, we introduce a benchmark, BETA, that aims to address this gap by (i) providing an extensive multipartite network consisting of 0.97 million biomedical concepts and 8.5 million associations, in addition to 62 million drug-drug and protein-protein similarities and (ii) presenting evaluation strategies that reflect seven cases (i.e. general, screening with different connectivity, target and drug screening based on categories, searching for specific drugs and targets and drug repurposing for specific diseases), a total of seven Tests (consisting of 344 Tasks in total) across multiple sampling and validation strategies. Six state-of-the-art methods covering two broad input data types (chemical structure- and gene sequence-based and network-based) were tested across all the developed Tasks. The best-worst performing cases have been analyzed to demonstrate the ability of the proposed benchmark to identify limitations of the tested methods for running over the benchmark tasks. The results highlight BETA as a benchmark in the selection of computational strategies for drug repurposing and target discovery.

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.

Exploring Novel Computable Knowledge in Structured Drug Product Labels

This paper introduces a database derived from Structured Product Labels (SPLs). SPLs are legally mandated snapshots containing information on all drugs released to market in the United States. Since publication is not required for pre-trial findings, we hypothesize that SPLs may contain knowledge absent in the literature, and hence "novel." SemMedDB is an existing database of computable knowledge derived from the literature. If SPL content could be similarly transformed, novel clinically relevant assertions in the SPLs could be identified through comparison with SemMedDB. After we derive a database (containing 4,297,481 assertions), we compare the extracted content with SemMedDB for recent FDA drug approvals. We find that novelty between the SPLs and the literature is nuanced, due to the redundancy of SPLs. Highlighting areas for improvement and future work, we conclude that SPLs contain a wealth of novel knowledge relevant to research and complementary to the literature.

Computational Advances in Drug Safety: Systematic and Mapping Review of Knowledge Engineering Based Approaches

Drug Safety (DS) is a domain with significant public health and social impact. Knowledge Engineering (KE) is the Computer Science discipline elaborating on methods and tools for developing “knowledge-intensive” systems, depending on a conceptual “knowledge” schema and some kind of “reasoning” process. The present systematic and mapping review aims to investigate KE-based approaches employed for DS and highlight the introduced added value as well as trends and possible gaps in the domain. Journal articles published between 2006 and 2017 were retrieved from PubMed/MEDLINE and Web of Science® (873 in total) and filtered based on a comprehensive set of inclusion/exclusion criteria. The 80 finally selected articles were reviewed on full-text, while the mapping process relied on a set of concrete criteria (concerning specific KE and DS core activities, special DS topics, employed data sources, reference ontologies/terminologies, and computational methods, etc.). The analysis results are publicly available as online interactive analytics graphs. The review clearly depicted increased use of KE approaches for DS. The collected data illustrate the use of KE for various DS aspects, such as Adverse Drug Event (ADE) information collection, detection, and assessment. Moreover, the quantified analysis of using KE for the respective DS core activities highlighted room for intensifying research on KE for ADE monitoring, prevention and reporting. Finally, the assessed use of the various data sources for DS special topics demonstrated extensive use of dominant data sources for DS surveillance, i.e., Spontaneous Reporting Systems, but also increasing interest in the use of emerging data sources, e.g., observational healthcare databases, biochemical/genetic databases, and social media. Various exemplar applications were identified with promising results, e.g., improvement in Adverse Drug Reaction (ADR) prediction, detection of drug interactions, and novel ADE profiles related with specific mechanisms of action, etc. Nevertheless, since the reviewed studies mostly concerned proof-of-concept implementations, more intense research is required to increase the maturity level that is necessary for KE approaches to reach routine DS practice. In conclusion, we argue that efficiently addressing DS data analytics and management challenges requires the introduction of high-throughput KE-based methods for effective knowledge discovery and management, resulting ultimately, in the establishment of a continuous learning DS system.

Learning Opportunities for Drug Repositioning via GWAS and PheWAS Findings

Drug repositioning for available medications can be preferred over traditional drug development, which requires substantially more effort to uncover new insights into medications and diseases. Genome-Wide Association Studies (GWAS) and Phenome-Wide Association Studies (PheWAS) are two complimentary methods for finding novel associations between genes and diseases. We hypothesize that discoveries from these studies could be leveraged to find new targets for existing drugs. Thus, we propose a framework to learn opportunities for inferring such relationships via overlapped genes between disease-associated genes (e.g. GWAS and PheWAS findings) and drugtargeted genes. We use drug indications found in Medication Indication Resource (MEDI) as a gold standard to evaluate if drug indications learned from GWAS and PheWAS findings have clinical indications. We examined 151,011 <drug, GWAS phenotype> pairs from 987 drugs across 153 diseases and 763 pairs were statistically significant. Out of these 763 pairs, 16 of them were found to have clinical indications.

Information needs for making clinical recommendations about potential drug-drug interactions: a synthesis of literature review and interviews

BACKGROUND

Drug information compendia and drug-drug interaction information databases are critical resources for clinicians and pharmacists working to avoid adverse events due to exposure to potential drug-drug interactions (PDDIs). Our goal is to develop information models, annotated data, and search tools that will facilitate the interpretation of PDDI information. To better understand the information needs and work practices of specialists who search and synthesize PDDI evidence for drug information resources, we conducted an inquiry that combined a thematic analysis of published literature with unstructured interviews.

METHODS

Starting from an initial set of relevant articles, we developed search terms and conducted a literature search. Two reviewers conducted a thematic analysis of included articles. Unstructured interviews with drug information experts were conducted and similarly coded. Information needs, work processes, and indicators of potential strengths and weaknesses of information systems were identified.

RESULTS

Review of 92 papers and 10 interviews identified 56 categories of information needs related to the interpretation of PDDI information including drug and interaction information; study design; evidence including clinical details, quality and content of reports, and consequences; and potential recommendations. We also identified strengths/weaknesses of PDDI information systems.

CONCLUSIONS

We identified the kinds of information that might be most effective for summarizing PDDIs. The drug information experts we interviewed had differing goals, suggesting a need for detailed information models and flexible presentations. Several information needs not discussed in previous work were identified, including temporal overlaps in drug administration, biological plausibility of interactions, and assessment of the quality and content of reports. Richly structured depictions of PDDI information may help drug information experts more effectively interpret data and develop recommendations. Effective information models and system designs will be needed to maximize the utility of this information.

Overlap in drug-disease associations between clinical practice guidelines and drug structured product label indications

BACKGROUND

Clinical practice guidelines (CPGs) recommend pharmacologic treatments for clinical conditions, and drug structured product labels (SPLs) summarize approved treatment indications. Both resources are intended to promote evidence-based medical practices and guide clinicians' prescribing decisions. However, it is unclear how well CPG recommendations about pharmacologic therapies match SPL indications for recommended drugs. In this study, we perform text mining of CPG summaries to examine drug-disease associations in CPG recommendations and in SPL treatment indications for 15 common chronic conditions.

METHODS

We constructed an initial text corpus of guideline summaries from the National Guideline Clearinghouse (NGC) from a set of manually selected ICD-9 codes for each of the 15 conditions. We obtained 377 relevant guideline summaries and their Major Recommendations section, which excludes guidelines for pediatric patients, pregnant or breastfeeding women, or for medical diagnoses not meeting inclusion criteria. A vocabulary of drug terms was derived from five medical taxonomies. We used named entity recognition, in combination with dictionary-based and ontology-based methods, to identify drug term occurrences in the text corpus and construct drug-disease associations. The ATC (Anatomical Therapeutic Chemical Classification) was utilized to perform drug name and drug class matching to construct the drug-disease associations from CPGs. We then obtained drug-disease associations from SPLs using conditions mentioned in their Indications section in SIDER. The primary outcomes were the frequency of drug-disease associations in CPGs and SPLs, and the frequency of overlap between the two sets of drug-disease associations, with and without using taxonomic information from ATC.

RESULTS

Without taxonomic information, we identified 1444 drug-disease associations across CPGs and SPLs for 15 common chronic conditions. Of these, 195 drug-disease associations overlapped between CPGs and SPLs, 917 associations occurred in CPGs only and 332 associations occurred in SPLs only. With taxonomic information, 859 unique drug-disease associations were identified, of which 152 of these drug-disease associations overlapped between CPGs and SPLs, 541 associations occurred in CPGs only, and 166 associations occurred in SPLs only.

CONCLUSIONS

Our results suggest that CPG-recommended pharmacologic therapies and SPL indications do not overlap frequently when identifying drug-disease associations using named entity recognition, although incorporating taxonomic relationships between drug names and drug classes into the approach improves the overlap. This has important implications in practice because conflicting or inconsistent evidence may complicate clinical decision making and implementation or measurement of best practices.

Using Nonexperts for Annotating Pharmacokinetic Drug-Drug Interaction Mentions in Product Labeling: A Feasibility Study

BACKGROUND

Because vital details of potential pharmacokinetic drug-drug interactions are often described in free-text structured product labels, manual curation is a necessary but expensive step in the development of electronic drug-drug interaction information resources. The use of nonexperts to annotate potential drug-drug interaction (PDDI) mentions in drug product label annotation may be a means of lessening the burden of manual curation.

OBJECTIVE

Our goal was to explore the practicality of using nonexpert participants to annotate drug-drug interaction descriptions from structured product labels. By presenting annotation tasks to both pharmacy experts and relatively naïve participants, we hoped to demonstrate the feasibility of using nonexpert annotators for drug-drug information annotation. We were also interested in exploring whether and to what extent natural language processing (NLP) preannotation helped improve task completion time, accuracy, and subjective satisfaction.

METHODS

Two experts and 4 nonexperts were asked to annotate 208 structured product label sections under 4 conditions completed sequentially: (1) no NLP assistance, (2) preannotation of drug mentions, (3) preannotation of drug mentions and PDDIs, and (4) a repeat of the no-annotation condition. Results were evaluated within the 2 groups and relative to an existing gold standard. Participants were asked to provide reports on the time required to complete tasks and their perceptions of task difficulty.

RESULTS

One of the experts and 3 of the nonexperts completed all tasks. Annotation results from the nonexpert group were relatively strong in every scenario and better than the performance of the NLP pipeline. The expert and 2 of the nonexperts were able to complete most tasks in less than 3 hours. Usability perceptions were generally positive (3.67 for expert, mean of 3.33 for nonexperts).

CONCLUSIONS

The results suggest that nonexpert annotation might be a feasible option for comprehensive labeling of annotated PDDIs across a broader range of drug product labels. Preannotation of drug mentions may ease the annotation task. However, preannotation of PDDIs, as operationalized in this study, presented the participants with difficulties. Future work should test if these issues can be addressed by the use of better performing NLP and a different approach to presenting the PDDI preannotations to users during the annotation workflow.

Toward a complete dataset of drug-drug interaction information from publicly available sources

Although potential drug-drug interactions (PDDIs) are a significant source of preventable drug-related harm, there is currently no single complete source of PDDI information. In the current study, all publically available sources of PDDI information that could be identified using a comprehensive and broad search were combined into a single dataset. The combined dataset merged fourteen different sources including 5 clinically-oriented information sources, 4 Natural Language Processing (NLP) Corpora, and 5 Bioinformatics/Pharmacovigilance information sources. As a comprehensive PDDI source, the merged dataset might benefit the pharmacovigilance text mining community by making it possible to compare the representativeness of NLP corpora for PDDI text extraction tasks, and specifying elements that can be useful for future PDDI extraction purposes. An analysis of the overlap between and across the data sources showed that there was little overlap. Even comprehensive PDDI lists such as DrugBank, KEGG, and the NDF-RT had less than 50% overlap with each other. Moreover, all of the comprehensive lists had incomplete coverage of two data sources that focus on PDDIs of interest in most clinical settings. Based on this information, we think that systems that provide access to the comprehensive lists, such as APIs into RxNorm, should be careful to inform users that the lists may be incomplete with respect to PDDIs that drug experts suggest clinicians be aware of. In spite of the low degree of overlap, several dozen cases were identified where PDDI information provided in drug product labeling might be augmented by the merged dataset. Moreover, the combined dataset was also shown to improve the performance of an existing PDDI NLP pipeline and a recently published PDDI pharmacovigilance protocol. Future work will focus on improvement of the methods for mapping between PDDI information sources, identifying methods to improve the use of the merged dataset in PDDI NLP algorithms, integrating high-quality PDDI information from the merged dataset into Wikidata, and making the combined dataset accessible as Semantic Web Linked Data.

Utilizing the Wikidata system to improve the quality of medical content in Wikipedia in diverse languages: a pilot study

Background

Wikipedia is an important source of medical information for both patients and medical professionals. Given its wide reach, improving the quality, completeness, and accessibility of medical information on Wikipedia could have a positive impact on global health.

Objective

We created a prototypical implementation of an automated system for keeping drug-drug interaction (DDI) information in Wikipedia up to date with current evidence about clinically significant drug interactions. Our work is based on Wikidata, a novel, graph-based database backend of Wikipedia currently in development.

Methods

We set up an automated process for integrating data from the Office of the National Coordinator for Health Information Technology (ONC) high priority DDI list into Wikidata. We set up exemplary implementations demonstrating how the DDI data we introduced into Wikidata could be displayed in Wikipedia articles in diverse languages. Finally, we conducted a pilot analysis to explore if adding the ONC high priority data would substantially enhance the information currently available on Wikipedia.

Results

We derived 1150 unique interactions from the ONC high priority list. Integration of the potential DDI data from Wikidata into Wikipedia articles proved to be straightforward and yielded useful results. We found that even though the majority of current English Wikipedia articles about pharmaceuticals contained sections detailing contraindications, only a small fraction of articles explicitly mentioned interaction partners from the ONC high priority list. For 91.30% (1050/1150) of the interaction pairs we tested, none of the 2 articles corresponding to the interacting substances explicitly mentioned the interaction partner. For 7.21% (83/1150) of the pairs, only 1 of the 2 associated Wikipedia articles mentioned the interaction partner; for only 1.48% (17/1150) of the pairs, both articles contained explicit mentions of the interaction partner.

Conclusions

Our prototype demonstrated that automated updating of medical content in Wikipedia through Wikidata is a viable option, albeit further refinements and community-wide consensus building are required before integration into public Wikipedia is possible. A long-term endeavor to improve the medical information in Wikipedia through structured data representation and automated workflows might lead to a significant improvement of the quality of medical information in one of the world’s most popular Web resources.

Automating Identification of Multiple Chronic Conditions in Clinical Practice Guidelines

Many clinical practice guidelines (CPGs) are intended to provide evidence-based guidance to clinicians on a single disease, and are frequently considered inadequate when caring for patients with multiple chronic conditions (MCC), or two or more chronic conditions. It is unclear to what degree disease-specific CPGs provide guidance about MCC. In this study, we develop a method for extracting knowledge from single-disease chronic condition CPGs to determine how frequently they mention commonly co-occurring chronic diseases. We focus on 15 highly prevalent chronic conditions. We use publicly available resources, including a repository of guideline summaries from the National Guideline Clearinghouse to build a text corpus, a data dictionary of ICD-9 codes from the Medicare Chronic Conditions Data Warehouse (CCW) to construct an initial list of disease terms, and disease synonyms from the National Center for Biomedical Ontology to enhance the list of disease terms. First, for each disease guideline, we determined the frequency of comorbid condition mentions (a disease-comorbidity pair) by exactly matching disease synonyms in the text corpus. Then, we developed an annotated reference standard using a sample subset of guidelines. We used this reference standard to evaluate our approach. Then, we compared the co-prevalence of common pairs of chronic conditions from Medicare CCW data to the frequency of disease-comorbidity pairs in CPGs. Our results show that some disease-comorbidity pairs occur more frequently in CPGs than others. Sixty-one (29.0%) of 210 possible disease-comorbidity pairs occurred zero times; for example, no guideline on chronic kidney disease mentioned depression, while heart failure guidelines mentioned ischemic heart disease the most frequently. Our method adequately identifies comorbid chronic conditions in CPG recommendations with precision 0.82, recall 0.75, and F-measure 0.78. Our work identifies knowledge currently embedded in the free text of clinical practice guideline recommendations and provides an initial view of the extent to which CPGs mention common comorbid conditions. Knowledge extracted from CPG text in this way may be useful to inform gaps in guideline recommendations regarding MCC and therefore identify potential opportunities for guideline improvement.

Integrative approaches for predicting in vivo effects of chemicals from their structural descriptors and the results of short-term biological assays

Cheminformatics approaches such as Quantitative Structure Activity Relationship (QSAR) modeling have been used traditionally for predicting chemical toxicity. In recent years, high throughput biological assays have been increasingly employed to elucidate mechanisms of chemical toxicity and predict toxic effects of chemicals in vivo. The data generated in such assays can be considered as biological descriptors of chemicals that can be combined with molecular descriptors and employed in QSAR modeling to improve the accuracy of toxicity prediction. In this review, we discuss several approaches for integrating chemical and biological data for predicting biological effects of chemicals in vivo and compare their performance across several data sets. We conclude that while no method consistently shows superior performance, the integrative approaches rank consistently among the best yet offer enriched interpretation of models over those built with either chemical or biological data alone. We discuss the outlook for such interdisciplinary methods and offer recommendations to further improve the accuracy and interpretability of computational models that predict chemical toxicity.

Toward semantic modeling of pharmacogenomic knowledge for clinical and translational decision support

The purpose of this paper is to describe pilot work on a semantic model of the pharmacogenomics information found in drug product labels. The model's development is driven by a series of use cases that we have developed to demonstrate how structured pharmacogenomics information could be more effectively used to support clinical and translational efforts. Using an iterative process, the semantic model was field-tested by five pharmacists, who used it to manually annotate a subset of the sections that the Food and Drug Administration's Table of Pharmacogenomic Biomarkers in Drug Labels cites as containing pharmacogenomics information. The five pharmacists identified a total of 213 pharmacogenomics statements in 29 sections. The model showed the potential to make the unstructured pharmacogenomic information currently written in product labeling more accessible and actionable through structured annotations of pharmacogenomics effects and clinical recommendations.

Extending the "web of drug identity" with knowledge extracted from United States product labels

Structured Product Labels (SPLs) contain information about drugs that can be valuable to clinical and translational research, especially if it can be linked to other sources that provide data about drug targets, chemical properties, interactions, and biological pathways. Unfortunately, SPLs currently provide coarsely-structured drug information and lack the detailed annotation that is required to support computational use cases. To help address this issue we created LinkedSPLs, a Linked Data resource that extends the "web of drug identity" using information extracted from SPLs. In this paper we describe the mapping that LinkedSPLs provides between SPL active ingredients and DrugBank chemical entities. These mappings were created using three approaches: InChI chemical structure descriptors comparison, exact string matching based on the chemical name, and automatic (unsupervised) linkage identification. Comparison of the approaches found that, while these three approaches are complementary, the automatic approach performs well in terms of precision and recall.