DOCSLIB.ORG

Reconstructing Manual Information Extraction with DB-To-Document Backprojection: Experiments in the Life Science Domain

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Reconstructing Manual Information Extraction with DB-To-Document Backprojection: Experiments in the Life Science Domain Reconstructing Manual Information Extraction with DB-to-Document Backprojection: Experiments in the Life Science Domain Mark-Christoph Muller,¨ Sucheta Ghosh, Maja Rey, Ulrike Wittig, Wolfgang Muller¨ and Michael Strube Heidelberg Institute for Theoretical Studies gGmbH, Heidelberg, Germany fmark-christoph.mueller,sucheta.ghosh,maja.rey, ulrike.wittig,wolfgang.mueller,[email protected] Abstract et al., 2019; Huang et al., 2020; Wu et al., 2020; Abdelhakim et al., 2020), because data quality (i.e. We introduce a novel scientific document pro- correctness and integrity) has priority over quan- cessing task for making previously inaccessi- tity (i.e. more quickly available, but potentially less ble information in printed paper documents reliable, data), and the error rates of current NLP available to automatic processing. We de- scribe our data set of scanned documents systems are still considered too high (Karp, 2016). and data records from the biological database For reasons of ergonomics and ease of handling SABIO-RK, provide a definition of the task, (Buchanan and Loizides, 2007;K opper¨ et al., 2016; and report findings from preliminary experi- Clinton, 2019), the identification and mark-up ments. Rigorous evaluation proved challeng- step often involves paper printouts and highlighter ing due to lack of gold-standard data and a dif- pens,2 like in the example page in Figure1. ficult notion of correctness. Qualitative inspec- tion of results, however, showed the feasibility and usefulness of the task. 1 Introduction Research results from the life sciences are mainly published in the form of written journal or confer- ence papers, even though these results often take the form of measurements of experimental parame- ters, which would more appropriately be stored in a structured, machine-readable form. While there is some tendency towards directly publishing ex- perimental data, e.g. on SourceData (Liechti et al., Figure 1: Page with mark-up (best viewed in color). 2016) or (for environmental data) PANGAEA1, this is not the norm yet, and does not help with the As mere intermediate products of the curation pro- huge body of data already published in the con- cess, the manually highlighted printouts are only ventional literature. It is common practice in the required until all data from the respective docu- life sciences, therefore, to manually extract infor- ment has been curated, and they will normally be mation (including measurements and the experi- archived afterwards. We argue, however, that the mental conditions underlying them) from natural printouts contain even more information which cu- language documents, and to use it to populate bio- ration simply does not make full use of: First, some logical databases. This process is called biocura- document sections, although containing highlight- tion (International Society for Biocuration, 2018) ing, will not lead to the creation of a record in the and comprises, for every document, 1) identifica- biological database (see our results in Section 5.3). tion and mark-up of curatable information, 2) data Yet, this highlighting can still be regarded as a kind extraction, normalization, and consolidation, and of relevance annotation, produced by life science 3) database insertion. Despite constant improve- 2This is true for our own group, and has been corroborated ments in NLP technology, biocuration involves sig- in 2016 by an informal, unpublished survey among 21 curators nificant human labor (mostly reading) (Oughtred from 15+ biological databases. The survey showed that a considerable number of curators rely on paper printouts for 1www.pangaea.de close reading and / or highlighting of important information. 81 Proceedings of the First Workshop on Scholarly Document Processing, pages 81–90 Online, November 19, 2020. c 2020 Association for Computational Linguistics https://doi.org/10.18653/v1/P17 domain experts through attentive, task-oriented version of our archive of 6; 000+ manually high- reading. Obviously, this information should be lighted printouts of documents from the life science useful, e.g. for the analysis of how important in- domain, which have been curated in the 10+ years formation is dispersed over a scientific document. of our database’s existence. Over the years, numer- Second, for those database records that are created ous different curators were involved in the manual from highlighted document sections, the reference mark-up. Different highlighter colors were used, to that section is normally not preserved. Again, sometimes even within the same document (see an obvious way to use this information is to allow Figure1). In case of equivalent information ap- users of the biological database to visually trace pearing repeatedly in the same document, curators the record to its source in the document, including generally attempted to be economical and to avoid the original context. redundancy by highlighting only the most appro- In this paper, we describe our approach towards priate appearance, which is often, but not always, re-purposing scientific document printouts which the first appearance. While the mark-up was per- were manually highlighted during biocuration. formed in a completely unrestricted manner (cf. More precisely, our research question is: Given below), in the vast majority of cases, highlighting records of curated information from the database was applied directly to words or lines (cf. Figure1), and the original, scanned source document, (to which greatly helped in extracting the highlighted what degree) can we recover the document sec- text (cf. Section3). In some rare cases, curators se- tion that a particular record was extracted from? lected whole sections by drawing a vertical line at We consider this to be a novel scientific document the section’s margin. Also, data in tables was some- processing task, and propose to refer to it as DB- times highlighted on the cell level, while in other to-document backprojection. The remainder of cases, only the column header, the table header, or the paper is structured as follows. In Section2 we even the table caption was highlighted. We cre- describe the data basis of our work. Section3 intro- ated an electronic version of the document collec- duces the highlighted text extraction task, which we tion by scanning and OCR-processing all papers4, consider as self-contained and only loosely linked which resulted in a sandwich PDF for each docu- to the main task. Section4 deals with the actual ment with the (partially highlighted) background DB-to-document backprojection task, provides a superimposed with the extracted text. OCR was precise definition, and describes our processing performed with commercial software (Alaris Cap- steps. Section5 presents some preliminary experi- ture Pro), which was used out-of-the-box. The total ments, results, and error analysis.Initially, this sec- number of tokens in the 98 documents is 630; 153, tion will also discuss our approach to evaluation. with 6; 430 tokens/document on average. In Section6 we discuss some related work, and The second data set is the record data set which Section7 contains our conclusions and directions contains measurements of kinetic parameters that for the future. Note that, although our data is from were extracted from individual documents from the life sciences, the task is relevant for all domains the document collection in the course of manual where manual information extraction is performed curation. Each of the 2; 916 records in this data on natural language documents (like e.g. in Lipani set is linked to exactly one source document (via et al.(2014), where information is extracted from its PubMed ID), but no lower-level links (to pages IR research papers in the form of machine-readable or lines) exist. Each document, in turn, can be ’nanopublications’). linked to an arbitrary number of records (29:76 records/document on average). It has to be noted 2 Data that the above count of 2; 916 records contains a The work in this paper is based on two related data considerable number of multiple counts. This is sets, which have been collected in the SABIO-RK true in particular for records of type experimental Biochemical Reaction Kinetics Database project3. condition (cf. below), and is due to the fact that SABIO-RK is a curated database containing struc- often, several measurements are performed under tured information about biochemical reactions and identical experimental conditions. For scoring and their corresponding kinetics (Wittig et al., 2017, evaluation, however, this does not make a differ- 2018). The document data set is an electronic 4For the experiments reported in this paper, we only use a 3http://sabio.h-its.org/ subset of 98 documents. 82 ence, because we conflate semantically identical consist of document pages from which pixels that records before analysis. There are two main types were detected as belonging to text were removed of records, experimental condition and parameter. by inpainting (see ’Page background image’ in the Each record consists of three to six attribute-value lower left part of Figure4). pdftotext, on the (a-v) pairs. Figures2 and3 show one example of other hand, is used to extract the text that was pre- each type of record. viously recognized by OCR. It produces one XML file for the document, incl. bounding boxes on the conditionName: ’pH’, token-level. These tokens reflect the original doc- startValue : 7.7, buffer : ’0.10 M Tris-HCl, ument layout, but come in correct reading order 100 mM KCl, 1 mM DTT, even for multi-column documents. The second step 4.0 mM MgCl2, makes use of some simple image processing. As 10% Glycerol’ described above (Section2), document highlight- Figure 2: Record of type experimental condition with ing can come in any color, so searching for areas of three a-v pairs featuring one numeric, one atomic string, any particular color (like e.g. yellow) is not an op- and one complex string value. tion. Instead, our algorithm combines the facts that 1) highlighting is always non-grey and 2) shades of grey in the RGB color model are characterized parameterName : ’Km’, by identical, or at least highly similar, values in unitName : ’µM’, 6 startValue : 123, the R, G, and B components.
Load more

Recommended publications
  • Original Article Text Mining in the Biocuration Workflow: Applications for Literature Curation at Wormbase, Dictybase and TAIR
    Database, Vol. 2012, Article ID bas040, doi:10.1093/database/bas040 ............................................................................................................................................................................................................................................................................................. Original article Text mining in the biocuration workflow: applications for literature curation at WormBase, dictyBase and TAIR Kimberly Van Auken1,*, Petra Fey2, Tanya Z. Berardini3, Robert Dodson2, Laurel Cooper4, Donghui Li3, Juancarlos Chan1, Yuling Li1, Siddhartha Basu2, Hans-Michael Muller1, Downloaded from Rex Chisholm2, Eva Huala3, Paul W. Sternberg1,5 and the WormBase Consortium 1Division of Biology, California Institute of Technology, 1200 E. California Boulevard, Pasadena, CA 91125, 2Northwestern University Biomedical Informatics Center and Center for Genetic Medicine, 420 E. Superior Street, Chicago, IL 60611, 3Department of Plant Biology, Carnegie Institution, 260 Panama Street, Stanford, CA 94305, 4Department of Botany and Plant Pathology, Oregon State University, Corvallis, OR 97331 and 5Howard Hughes Medical Institute, California Institute of Technology, 1200 E. California Boulevard, Pasadena, CA 91125, USA http://database.oxfordjournals.org/ *Corresponding author: Tel: +1 609 937 1635; Fax: +1 626 568 8012; Email: [email protected] Submitted 18 June 2012; Revised 30 September 2012; Accepted 2 October 2012 ............................................................................................................................................................................................................................................................................................
    [Show full text]
  • Biocuration 2016 - Posters
    Biocuration 2016 - Posters Source: http://www.sib.swiss/events/biocuration2016/posters 1 RAM: A standards-based database for extracting and analyzing disease-specified concepts from the multitude of biomedical resources Jinmeng Jia and Tieliu Shi Each year, millions of people around world suffer from the consequence of the misdiagnosis and ineffective treatment of various disease, especially those intractable diseases and rare diseases. Integration of various data related to human diseases help us not only for identifying drug targets, connecting genetic variations of phenotypes and understanding molecular pathways relevant to novel treatment, but also for coupling clinical care and biomedical researches. To this end, we built the Rare disease Annotation & Medicine (RAM) standards-based database which can provide reference to map and extract disease-specified information from multitude of biomedical resources such as free text articles in MEDLINE and Electronic Medical Records (EMRs). RAM integrates disease-specified concepts from ICD-9, ICD-10, SNOMED-CT and MeSH (http://www.nlm.nih.gov/mesh/MBrowser.html) extracted from the Unified Medical Language System (UMLS) based on the UMLS Concept Unique Identifiers for each Disease Term. We also integrated phenotypes from OMIM for each disease term, which link underlying mechanisms and clinical observation. Moreover, we used disease-manifestation (D-M) pairs from existing biomedical ontologies as prior knowledge to automatically recognize D-M-specific syntactic patterns from full text articles in MEDLINE. Considering that most of the record-based disease information in public databases are textual format, we extracted disease terms and their related biomedical descriptive phrases from Online Mendelian Inheritance in Man (OMIM), National Organization for Rare Disorders (NORD) and Orphanet using UMLS Thesaurus.
    [Show full text]
  • Biocuration - Mapping Resources and Needs [Version 2; Peer Review: 2 Approved]
    F1000Research 2020, 9(ELIXIR):1094 Last updated: 22 JUL 2021 RESEARCH ARTICLE Biocuration - mapping resources and needs [version 2; peer review: 2 approved] Alexandra Holinski 1, Melissa L. Burke 1, Sarah L. Morgan 1, Peter McQuilton 2, Patricia M. Palagi 3 1European Molecular Biology Laboratory, European Bioinformatics Institute (EMBL-EBI), Hinxton, Cambridgeshire, CB10 1SD, UK 2Oxford e-Research Centre, Department of Engineering Science, University of Oxford, Oxford, Oxfordshire, OX1 3QG, UK 3SIB Swiss Institute of Bioinformatics, Lausanne, 1015, Switzerland v2 First published: 04 Sep 2020, 9(ELIXIR):1094 Open Peer Review https://doi.org/10.12688/f1000research.25413.1 Latest published: 02 Dec 2020, 9(ELIXIR):1094 https://doi.org/10.12688/f1000research.25413.2 Reviewer Status Invited Reviewers Abstract Background: Biocuration involves a variety of teams and individuals 1 2 across the globe. However, they may not self-identify as biocurators, as they may be unaware of biocuration as a career path or because version 2 biocuration is only part of their role. The lack of a clear, up-to-date (revision) report profile of biocuration creates challenges for organisations like ELIXIR, 02 Dec 2020 the ISB and GOBLET to systematically support biocurators and for biocurators themselves to develop their own careers. Therefore, the version 1 ELIXIR Training Platform launched an Implementation Study in order 04 Sep 2020 report report to i) identify communities of biocurators, ii) map the type of curation work being done, iii) assess biocuration training, and iv) draw a picture of biocuration career development. 1. Tanya Berardini , Phoenix Bioinformatics, Methods: To achieve the goals of the study, we carried out a global Fremont, USA survey on the nature of biocuration work, the tools and resources that are used, training that has been received and additional training 2.
    [Show full text]
  • Improving the Gene Ontology Resource to Facilitate More Informative Analysis and Interpretation of Alzheimer’S Disease Data
    G C A T T A C G G C A T genes Article Improving the Gene Ontology Resource to Facilitate More Informative Analysis and Interpretation of Alzheimer’s Disease Data Barbara Kramarz 1 , Paola Roncaglia 2 , Birgit H. M. Meldal 2 , Rachael P. Huntley 1 , Maria J. Martin 2, Sandra Orchard 2, Helen Parkinson 2, David Brough 3, Rina Bandopadhyay 4, Nigel M. Hooper 3 and Ruth C. Lovering 1,* 1 UCL Institute of Cardiovascular Science, University College London, Rayne Building, 5 University Street, London WC1E 6JF, UK; [email protected] (B.K.); [email protected] (R.P.H.) 2 European Bioinformatics Institute (EMBL-EBI), European Molecular Biology Laboratory, Wellcome Genome Campus, Hinxton, Cambridge CB10 1SD, UK; [email protected] (P.R.); [email protected] (B.H.M.M.); [email protected] (M.J.M.); [email protected] (S.O.); [email protected] (H.P.) 3 Division of Neuroscience and Experimental Psychology, School of Biological Sciences, Faculty of Biology, Medicine and Health, Manchester Academic Health Science Centre, University of Manchester, AV Hill Building, Oxford Road, Manchester M13 9PT, UK; [email protected] (D.B.); [email protected] (N.M.H.) 4 UCL Queen Square Institute of Neurology and Reta Lila Weston Institute of Neurological Studies, 1 Wakefield Street, London WC1N 1PJ, UK; [email protected] * Correspondence: [email protected] or [email protected]; Tel.: +44-207-679-6965 Received: 31 October 2018; Accepted: 23 November 2018; Published: 29 November 2018 Abstract: The analysis and interpretation of high-throughput datasets relies on access to high-quality bioinformatics resources, as well as processing pipelines and analysis tools.
    [Show full text]
  • Biocuration Experts on the Impact of Duplication and Other Data Quality Issues in Biological Databases
    Genomics Proteomics Bioinformatics 18 (2020) 91–103 Genomics Proteomics Bioinformatics www.elsevier.com/locate/gpb www.sciencedirect.com PERSPECTIVE Quality Matters: Biocuration Experts on the Impact of Duplication and Other Data Quality Issues in Biological Databases Qingyu Chen 1,*, Ramona Britto 2, Ivan Erill 3, Constance J. Jeffery 4, Arthur Liberzon 5, Michele Magrane 2, Jun-ichi Onami 6,7, Marc Robinson-Rechavi 8,9, Jana Sponarova 10, Justin Zobel 1,*, Karin Verspoor 1,* 1 School of Computing and Information Systems, University of Melbourne, Melbourne, VIC 3010, Australia 2 European Molecular Biology Laboratory, European Bioinformatics Institute (EMBL-EBI), Wellcome Trust Genome Campus, Cambridge CB10 1SD, UK 3 Department of Biological Sciences, University of Maryland, Baltimore, MD 21250, USA 4 Department of Biological Sciences, University of Illinois at Chicago, Chicago, IL 60607, USA 5 Broad Institute of MIT and Harvard, Cambridge, MA 02142, USA 6 Japan Science and Technology Agency, National Bioscience Database Center, Tokyo 102-8666, Japan 7 National Institute of Health Sciences, Tokyo 158-8501, Japan 8 Swiss Institute of Bioinformatics, CH-1015 Lausanne, Switzerland 9 Department of Ecology and Evolution, University of Lausanne, CH-1015 Lausanne, Switzerland 10 Nebion AG, 8048 Zurich, Switzerland Received 8 December 2017; revised 24 October 2018; accepted 14 December 2018 Available online 9 July 2020 Handled by Zhang Zhang Introduction assembled, annotated, and ultimately submitted to primary nucleotide databases such as GenBank [2], European Nucleo- tide Archive (ENA) [3], and DNA Data Bank of Japan Biological databases represent an extraordinary collective vol- (DDBJ) [4] (collectively known as the International Nucleotide ume of work.
    [Show full text]
  • ISB Newsletter
    ISB June 2012 . INTERNATIONAL SOCIETY ISB FOR BIOCURATION June, 2012 In this issue: - Save the Date: ISB2013 Save the Date - B3CB Workshop Report - ISB Membership Renewal for ISB2013! - 2012 Elections: ISB Executive Committee The 6th International Biocuration Conference - Jack Leunissen will be held on April 7-10, - Latest Advances from ISB Research Teams 2013, at Churchill College, - Job Opportunities Cambridge, UK. Training Activities: B3CB Workshop Report B3CB was a workshop aimed at developing common resources and infrastructures for training of experts as well as users of bioinformatics tools and resources. It was held in Uppsala (Sweden), and chaired by Terri Attwood (Manchester University, ISB Executive Committee member). Pascale Gaudet introduced the ISB. We believe that this initiative will help the ISB provide more support for training, a part of the our mission statement. The workshop brought together members of the executive committees of several ISB Membership Renewal organizations in Bioinformatics, Biotechnology, Biocuration and Computational Biology (B3CB): For many of us, membership fees are due - EMBnet (Global Bioinformatics Network) – Terri Attwood this month! Notices for renewal of ISB - ISCB (International Society for Computational Biology) – Reinhard Schneider Memberships will be reaching inboxes in - APBioNet (Asia-Pacific Bioinformatics Network) – Christian Schönbach the coming days. Your ISB Membership - ASBCB (African Society for Bioinformatics & Computational Biology) – Nicky Mulder goes to support activities such as the - SoIBio (Iberoamerican Society for Bioinformatics) – Oswaldo Trelles International Biocuration Conference - ISB (International Society for Biocuration) – Pascale Gaudet (travel support was provided for 10 - EBI/ELIXIR (European Bioinformatics Institute) – Cath Brooksbank attendees to ISB2012), BioDBCore (in - NBIC (Netherlands Bioinformatics Centre) – Celia van Gelder collaboration with BioSharing), and to - SeqAhead (Next-Gen Sequencing Data Analysis Network) – Erik Bongcam-Rudloff establish links with other groups.
    [Show full text]
  • Ontology: Tool for Broad Spectrum Knowledge Integration Barry Smith
    Foreword to Chinese Translation Ontology: Tool for Broad Spectrum Knowledge Integration Barry Smith BFO: The Beginnings This book was first published in 2015. Its primary target audience was bio- and biomedical informaticians, reflecting the ways in which ontologies had become an established part of the toolset of bio- and biomedical informatics since (roughly) the completion of the Human Genome Project (HGP). As is well known, the success of HGP led to the transformation of biological and clinical sciences into information-driven disciplines and spawned a whole series of new disciplines with names like ‘proteomics’, ‘connectomics’ and ‘toxiocopharmacogenomics’. It was of course not only the human genome that was made available for research but also the genomes of other ‘model organisms’, such as mouse or fly. The remarkable similarities between these genomes and the human genome made it possible to carry out experiments on model organisms and use the results to draw conclusions relevant to our understanding of human health and disease. To bring this about, however, it was necessary to create a controlled vocabulary that could be used for describing salient features of model organisms in a species-neutral way, and to use the terms of this vocabulary to tag the sequence data for all salient organisms. It was with the purpose of creating such a vocabulary that the Gene Ontology (GO) was born in a Montreal hotel bar in 1998.1 Since then the GO has served as mediator between the new genomic data on the one hand, which is accessible only with the aid of computers, and what we might think of as the ‘old biology data’ captured using natural language by means of terms such as ‘cell division’ or ‘mitochondrion’ or ‘protein binding’.
    [Show full text]
  • Integration of Proteomics Data Into Uniprotkb
    Benoit Bely1, Emanuele Alpi1,Guoying Qi1, Alan da Silva1, Jie Luo1, Maria Martin1 and the UniProt Consortium1, 2, 3 1 EMBL-European Bioinformatics Institute, Cambridge, UK 2 Swiss Institute of Bioinformatics, Geneva, Switzerland 3 Protein Information Resource, Washington DC, USA The Universal Protein Resource EBI is an outstation of the Integration of proteomics data European Molecular Biology Laboratory into UniProtKB Background - The amount of publicly available data in mass spectrometry (MS) proteomics repositories like PRIDE, PeptideAtlas, MaxQB, EPD, CTDP, ProteomicsDB, GPMDB, HPM and others, as well as data from third-party global PRIDE reprocessing, is growing at a fast rate. - The Universal Protein Resource (UniProt, http://www.uniprot.org/) is integrating proteomics data from all available public sources to provide comprehensive protein data sets with high quality experimentally proven evidences. UniProt protein existence (PE) values (http://www.uniprot.org/manual/protein_existence) >sp|Q8NI35|INADL_HUMAN InaD-like protein OS=Homo sapiens GN=INADL PE=1 SV=3 UniProt release 2016_02 1. Evidence at protein level PE value Nr of sequences PE value Nr of PE value Nr of (all species) (canonical only) % (human, reference proteome) sequences % (C. elegans , reference proteome) sequences % 2. Evidence at transcript level 1. Evidence at protein level 216550 0.35 1. Evidence at protein level 69831 75.77 1. Evidence at protein level 11414 41.14 2. Evidence at transcript level 1064707 1.73 2. Evidence at transcript level 7782 8.44 2. Evidence at transcript level 1091 3.93 3. Inferred from homology 3. Inferred from homology 13651703 22.19 3. Inferred from homology 1251 1.36 3.
    [Show full text]
  • Biocuration at the Saccharomyces Genome Database
    VC 2015 Wiley Periodicals, Inc. genesis 53:450–457 (2015) REVIEW Biocuration at the Saccharomyces Genome Database Marek S. Skrzypek,* and Robert S. Nash Department of Genetics, Saccharomyces Genome Database, Stanford University, Stanford, California Received 3 April 2015; Revised 12 May 2015; Accepted 13 May 2015 Summary: Saccharomyces Genome Database is an crosses and biochemical methods were being devel- online resource dedicated to managing information oped (Mortimer, 2000). In the decades that followed, about the biology and genetics of the model organism, our understanding of biochemical pathways and other yeast (Saccharomyces cerevisiae). This information is key aspects of cell biology, such as cell cycle control, derived primarily from scientific publications through a differentiation, and DNA repair were greatly informed process of human curation that involves manual extrac- by yeast research (Mortimer and Johnston, 1986). The tion of data and their organization into a comprehen- sive system of knowledge. This system provides a further development of molecular techniques, along foundation for further analysis of experimental data with the ease and relatively low cost of doing yeast coming from research on yeast as well as other organ- research, resulted in the coining of the phrase “The isms. In this review we will demonstrate how biocura- Awesome Power of Yeast Genetics” and elevated the tion and biocurators add a key component, the status of yeast as the premier model organism for the biological context, to our understanding of how genes, study of eukaryotic molecular and cellular biology proteins, genomes and cells function and interact. We (Duina et al., 2014). As a consequence of these advan- will explain the role biocurators play in sifting through ces in technology and the vast amount of knowledge the wealth of biological data to incorporate and con- that had accumulated, yeast became the best-studied nect key information.
    [Show full text]
  • Perspective on Literature Curation
    Mol Genet Genomics (2010) 283:415–425 DOI 10.1007/s00438-010-0525-8 REVIEW A MOD(ern) perspective on literature curation Jodi Hirschman • Tanya Z. Berardini • Harold J. Drabkin • Doug Howe Received: 4 November 2009 / Accepted: 6 February 2010 / Published online: 11 March 2010 Ó The Author(s) 2010. This article is published with open access at Springerlink.com Abstract Curation of biological data is a multi-faceted MODs to address: (1) the decision process by which papers task whose goal is to create a structured, comprehensive, are selected, and (2) the identification and prioritization of integrated, and accurate resource of current biological the data contained in the paper. We will highlight some of knowledge. These structured data facilitate the work of the the challenges that MOD biocurators face, and point to scientific community by providing knowledge about genes ways in which researchers and publishers can support the or genomes and by generating validated connections work of biocurators and the value of such support. between the data that yield new information and stimulate new research approaches. For the model organism dat- Keywords Annotation Á Biocuration Á Database Á abases (MODs), an important source of data is research Genome Á Literature Á Model organism publications. Every published paper containing experi- mental information about a particular model organism is a candidate for curation. All such papers are examined Introduction carefully by curators for relevant information. Here, four curators from different MODs describe the literature Biocuration, the practice of collecting and organizing curation process and highlight approaches taken by the four biological data, has roots that go back several thousand years.
    [Show full text]
  • OM2017 Camerafinal
    Exploring the Synergies between Biocuration and Ontology Alignment Automation David Dearing and Terrance Goan Stottler Henke Associates, Inc. 1107 NE 45th St, Suite 310 Seattle, WA 98105, USA {ddearing, goan}@stottlerhenke.com Abstract. Researchers have long recognized the value trapped in natural lan- guage publications and have continued to advance the development of ontologies that can help unleash this value. Among these advances are efforts to apply NLP techniques to streamline the labor-intensive process of scientific literature cura- tion, which encodes relevant information in a form that is accessible to both hu- mans and computers. In this paper, we report on our initial efforts to improve ontology alignment within the context of scientific literature curation by exploit- ing value within a large corpus of annotated PubMed abstracts. We employ an ensemble learning approach to augment a collection of publicly available ontol- ogy matching systems with a matching technique that leverages the word embed- dings learned from this corpus in order to more successfully match the concepts of two disease ontologies (MeSH and OMIM). Our experiments show that word embedding-based similarity scores do contribute value beyond traditional match- ing systems. Our results show that the performance of an ensemble trained on a small number of manually reviewed mappings is improved by their inclusion. Keywords: Ontology Matching Ensembles, Word Embeddings, Biocuration. 1 Introduction Technological advancements have given rise to an explosion in the rate that biomedical data is generated. The incredible volume of data now far exceeds the ability of re- searchers to capitalize on it. This is due, in large part, to the vagaries of the natural languages in which that data is published for consumption by human readers.
    [Show full text]
  • Building Curation Filters for BCO and Adapting the BCO Framework for Biocuration
    Building Curation Filters for BCO and Adapting the BCO Framework for Biocuration by Ngozi Onyegbula M.B.B.S. in Medicine and Surgery, July 1995, Nnamdi Azikiwe University A Thesis submitted to The Faculty of The Columbian College of Arts and Sciences of The George Washington University in partial fulfillment of the requirements for the degree of Master of Science August 31, 2020 Thesis directed by Jonathon Keeney Assistant Research Professor of Biochemistry and Molecular Medicine © Copyright 2020 by Ngozi Onyegbula All rights reserved ii Dedication This work is dedicated to my dear husband, Dr. Festus Onyegbula. My lovely children; Uzoma Onyegbula, Ugonna Onyegbula, and Chigozie Onyegbula. My amazing Mother, Rose Nwadiugwu. My soul sister, Amaka Nwadiugwu, and my precious nephew Ebube Onokpite. I thank God for all of you and your unwavering, unconditional love and support throughout the duration of my program. iii Acknowledgments I wish to acknowledge the following people for their unconditional support • Jonathon Keeney, PhD- My thesis adviser, always available to guide me. • Anelia Horvath, PhD- My reader who was there for me, despite late request. • Jack Vanderhoek, PhD- My program director. • Raja Mazumder, PhD- My course director. • Hayley Dingerdissen- For your advice. • Amanda Bell- Special thanks for all the support and encouragements. • BioCompute Lab Team- Especially Janisha Patel for your advice. • HIVE Lab Team. iv Abstract Building Curation Filters for BCO and Adapting the BCO Framework for Biocuration Biomedical research is becoming more interdisciplinary, and data sets are becoming increasingly large and complex. Many current methods in research (omics-type analyses) generate large quantities of data, and the internet generally makes access easy.
    [Show full text]