Dataset Information

DLAD4U: deriving and prioritizing disease lists from PubMed literature.

ABSTRACT:

Background

Due to recent technology advancements, disease related knowledge is growing rapidly. It becomes nontrivial to go through all published literature to identify associations between human diseases and genetic, environmental, and life style factors, disease symptoms, and treatment strategies. Here we report DLAD4U (Disease List Automatically Derived For You), an efficient, accurate and easy-to-use disease search engine based on PubMed literature.

Results

DLAD4U uses the eSearch and eFetch APIs from the National Center for Biotechnology Information (NCBI) to find publications related to a query and to identify diseases from the retrieved publications. The hypergeometric test was used to prioritize identified diseases for displaying to users. DLAD4U accepts any valid queries for PubMed, and the output results include a ranked disease list, information associated with each disease, chronologically-ordered supporting publications, a summary of the run, and links for file export. DLAD4U outperformed other disease search engines in our comparative evaluation using selected genes and drugs as query terms and manually curated data as "gold standard". For 100 genes that are associated with only one disease in the gold standard, the Mean Average Precision (MAP) measure from DLAD4U was 0.77, which clearly outperformed other tools. For 10 genes that are associated with multiple diseases in the gold standard, the mean precision, recall and F-measure scores from DLAD4U were always higher than those from other tools. The superior performance of DLAD4U was further confirmed using 100 drugs as queries, with an MAP of 0.90.

Conclusions

DLAD4U is a new, intuitive disease search engine that takes advantage of existing resources at NCBI to provide computational efficiency and uses statistical analyses to ensure accuracy. DLAD4U is publicly available at http://dlad4u.zhang-lab.org .

SUBMITTER: Shen J

PROVIDER: S-EPMC6309061 | biostudies-literature | 2018 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

DLAD4U: deriving and prioritizing disease lists from PubMed literature.

Shen Junhui J Vasaikar Suhas S Zhang Bing B

BMC bioinformatics 20181228 Suppl 17

<h4>Background</h4>Due to recent technology advancements, disease related knowledge is growing rapidly. It becomes nontrivial to go through all published literature to identify associations between human diseases and genetic, environmental, and life style factors, disease symptoms, and treatment strategies. Here we report DLAD4U (Disease List Automatically Derived For You), an efficient, accurate and easy-to-use disease search engine based on PubMed literature.<h4>Results</h4>DLAD4U uses the eSe ...[more]

PMID: 30591010

Similar Datasets

Project description:Answering questions such as "Which genes are related to breast cancer?" usually requires retrieving relevant publications through the PubMed search engine, reading these publications, and creating gene lists. This process is not only time-consuming, but also prone to errors.We report GLAD4U (Gene List Automatically Derived For You), a new, free web-based gene retrieval and prioritization tool. GLAD4U takes advantage of existing resources of the NCBI to ensure computational efficiency. The quality of gene lists created by GLAD4U for three Gene Ontology (GO) terms and three disease terms was assessed using corresponding "gold standard" lists curated in public databases. For all queries, GLAD4U gene lists showed very high recall but low precision, leading to low F-measure. As a comparison, EBIMed's recall was consistently lower than GLAD4U, but its precision was higher. To present the most relevant genes at the top of a list, we studied two prioritization methods based on publication count and the hypergeometric test, and compared the ranked lists and those generated by EBIMed to the gold standards. Both GLAD4U methods outperformed EBIMed for all queries based on a variety of quality metrics. Moreover, the hypergeometric method allowed for a better performance by thresholding genes with low scores. In addition, manual examination suggests that many false-positives could be explained by the incompleteness of the gold standards. The GLAD4U user interface accepts any valid queries for PubMed, and its output page displays the ranked gene list and information associated with each gene, chronologically-ordered supporting publications, along with a summary of the run and links for file export and functional enrichment and protein interaction network analysis.GLAD4U has a high overall recall. Although precision is generally low, the prioritization methods successfully rank truly relevant genes at the top of the lists to facilitate efficient browsing. GLAD4U is simple to use, and its interface can be found at: http://bioinfo.vanderbilt.edu/glad4u.

Project description:Active management of existing health technologies (e.g., devices, diagnostic, and/or medical procedures) to ensure the delivery of high value care is increasingly recognized around the world. A number of initiatives have raised awareness of technologies that may be overused, mis-used, or potentially harmful by compiling them into lists of low value care. However, despite the growing number of lists, changes to local healthcare practices remain challenging for many systems. The objective of this study was to develop and implement a process, leveraging existing initiatives and data assets, to produce a list of prioritized low value technologies for health technology reassessment (HTR).An expert advisory committee comprised of clinical experts and health system decision-makers was convened to determine key process requirements. Once developed, the process was piloted to assess feasibility in the Canadian province of British Columbia (BC).The expert advisory committee identified five required attributes for the process: data-driven, routine and replicable, actionable, stakeholder collaboration, and high return on investment. Guided by these attributes, a 5-step process was developed. First, over 1300 published low value technologies (i.e., from the National Institute for Health and Care Excellence [NICE] "do not do" recommendations, low value technologies in the Australian Medical Benefits Schedule, and Choosing Wisely "Top 5" lists) were identified. Using appropriate coding systems for BC's administrative health data (e.g., International Classification of Diseases [ICD]), the low value technologies were queried to examine frequencies and costs of technology use. This information was used to rank potential candidates for reassessment based on high annual budgetary impact. Lastly, clinical experts reviewed the ranked technologies prior to broad dissemination and stakeholder action. Pilot testing of the process in BC resulted in the prioritization of 9 initial candidate technologies for reassessment.This is the first account of a systematic approach to move a collective body of low value technology recommendations into action in a healthcare setting. This work demonstrates the feasibility and strength of using administrative data to identify and prioritize low value technologies for HTR at a population-level.

Project description:Background Clinical research publications have become the dominant source and basis of clinical evidence-based decision-making. Exploring the type and quantity of clinical research publications in the PubMed database is useful for clarifying the changing trends of clinical research development in recent years. Therefore, a longitudinal analysis of the type and quantity of clinical research publications in the PubMed database over three decades was conducted. Methods The PubMed database was searched to retrieve clinical research according to the type and year of publication from January 1, 1991 to December 31, 2020. The research types were classified as primary and secondary literature. Results A total of 1,078,404 primary literatures were retrieved and the constituent proportions were ranked from high to low as case report/series (27.54%), randomized clinical trials (RCTs) (23.62%), cohort studies (21.05%), cross-sectional studies (17.49%), case control studies (9.15%), non-RCTs (1.01%), and pragmatic clinical trials (PCTs) (0.15%). Correspondingly, 1,302,173 secondary literatures were retrieved and ranked as narrative review (70.88%), systematic review (15.02%), systematic review and meta-analyses (13.89%), traditional meta-analyses (4.48%), expert consensus (2.31%), guidelines (1.49%), scoping reviews (0.68%), net meta-analyses (0.40%), and umbrella reviews (0.04%). The average annual growth rate for the primary literature was 10.28%, and ranked from high to low as PCTs (83.68%), cohort studies (17.74%), cross-sectional studies (17.61%), non-RCTs (12.11%), case control studies (8.86%), RCTs (7.68%), case report/series (7.51%); while that for the secondary literature was 10.57%, and ranked from high to low as net meta-analyses (48.97%), umbrella reviews (47.09%), scoping reviews (41.92%), systematic reviews and meta-analyses (33.44%), systematic reviews (33.05%), traditional meta-analyses (12.49%), expert consensuses (9.22%), narrative review (8.72%), and guidelines (2.82%). Conclusion Both the composition and number of clinical studies changed significantly from 1991 to 2020. Based on the trend, the case report/series, case control study, and narrative review are on the decline, while cohort study, cross-sectional study, systematic reviews, and systematic review and meta-analysis literature have increased. To improve the quality of clinical evidence, we recommend RCT and cohort study give priority to access to allocated research resources in future. Supplementary Information The online version contains supplementary material available at 10.1186/s40001-022-00717-9.

Dataset Information

DLAD4U: deriving and prioritizing disease lists from PubMed literature.

Background

Results

Conclusions

Publications

DLAD4U: deriving and prioritizing disease lists from PubMed literature.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets