Assessing Immunology Dataset Quality through Literature Mining

The reproducibility issue of medical research is an important concern raised in the science community. Thecontinuing research in immunology keeps releasing massive and heterogeneous datasets. This unpredictedgrowth has led to challenges in finding and reusing these datasets. A comprehensive informatics frameworkto assess the immunology datasets is an immediate need. To respond to NIH/NIAID 147– Software or WebServices to Assess Quality and Reproducibility of Data and Information about Therapeutics and Vaccines,Short title: Software or web services to assess quality and reproducibility of data Fast Track Proposals will beaccepted, the research objective for this proposal, under the SBIR program, is to develop a comprehensiveinformatics framework to assess the quality of datasets available in the ImmPort data repository usingadvanced NLP methods. To achieve our goal, we propose the following two aims: (1) developing informationextraction approaches to recognize immunology dataset mentions in articles and link them to dataset IDs inthe ImmPort and (2) identifying context related to data quality for dataset mentions in articles using sentimentanalysis and generating aggregate data quality scores for each dataset. The successful completion of thisproject will provide a data quality framework for assessing datasets quality.