Browsing by Author "Zhang, Gongbo"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
Item Extracting descriptions of evidence for drug-drug interactions from the biomedical literature(University of Delaware, 2021) Zhang, GongboDrug-drug interactions (DDIs) can cause adverse drug effects, which result in morbidity and mortality. Predicting, anticipating, and avoiding them is thus an area of much concern. Studies of DDIs are typically published in the biomedical literature, however, gathering DDI information remains a challenging task given the rapidly increasing publication volume. In this thesis, we present a supervised learning approach to automatically collect evidence for DDIs from the biomedical literature. ☐ First, we identify the publications relevant to DDIs. The number of DDI relevant publications is relatively small compared to the vast amount of irrelevant publications. Models that attempt to identify publications relevant to DDIs without handling this imbalance are generally biased toward the irrelevant class, as the classifiers are trained on a dataset where most examples are irrelevant to DDIs. We propose a two-stage method to tackle this challenge. ☐ Potential DDIs can be verified with higher confidence if supporting evidence is provided. We thus focus on the descriptions of the evidence. Such evidence is typically described in statements conveying information about experimental methods and results. Analyzing such sentences can form the basis for effective retrieval of DDI knowledge. To identify the sentences containing evidential information, we introduce two methods based on statistical learning models and on neural networks. ☐ Different types of evidence for DDIs are reported in the biomedical literature. In vitro studies of drug interactions are conducted outside the living body, typically using animal or human cells, or material formed by recombination. In vivo studies focus on the impact of DDIs on living organisms. The drug interactions can also be discovered in clinical studies, which are used to check how well experimental treatments work as compared to the standard treatment for the same disease. Lack of any type of evidence is a knowledge gap that can hinder the process of transforming biomedical knowledge into clinical practice. Determining the type of evidence for DDIs forms a preliminary step toward systematically detecting and closing the gaps across the three types of evidence. We propose a supervised learning approach to identify the type of evidence for DDIs described in biomedical abstracts.Item Utilizing image and caption information for biomedical document classification(Bioinformatics, 2021-07-12) Li, Pengyuan; Jiang, Xiangying; Zhang, Gongbo; Trabucco, Juan Trelles; Raciti, Daniela; Smith, Cynthia; Ringwald, Martin; Marai, G. Elisabeta; Arighi, Cecilia; Shatkay, HagitMotivation: Biomedical research findings are typically disseminated through publications. To simplify access to domain-specific knowledge while supporting the research community, several biomedical databases devote significant effort to manual curation of the literature—a labor intensive process. The first step toward biocuration requires identifying articles relevant to the specific area on which the database focuses. Thus, automatically identifying publications relevant to a specific topic within a large volume of publications is an important task toward expediting the biocuration process and, in turn, biomedical research. Current methods focus on textual contents, typically extracted from the title-and-abstract. Notably, images and captions are often used in publications to convey pivotal evidence about processes, experiments and results. Results: We present a new document classification scheme, using both image and caption information, in addition to titles-and-abstracts. To use the image information, we introduce a new image representation, namely Figure-word, based on class labels of subfigures. We use word embeddings for representing captions and titles-and-abstracts. To utilize all three types of information, we introduce two information integration methods. The first combines Figure-words and textual features obtained from captions and titles-and-abstracts into a single larger vector for document representation; the second employs a meta-classification scheme. Our experiments and results demonstrate the usefulness of the newly proposed Figure-words for representing images. Moreover, the results showcase the value of Figure-words, captions and titles-and-abstracts in providing complementary information for document classification; these three sources of information when combined, lead to an overall improved classification performance. Availability and implementation: Source code and the list of PMIDs of the publications in our datasets are available upon request.