Negative findings can be hard to come by in the biomedical literature. And papers reporting such findings can be darn near impossible to turn up in searches of the go-to database, PubMed. Enter University of Wisconsin, Milwaukee, biomedical informatician Hong Yu. In collaboration with her grad student Shashank Agarwal, Yu developed a search engine, called BioNOT, that can find the rare papers reporting negative findings—such as a particular gene not being associated with a particular disease—that actually get published.
BioNOT uses artificial intelligence and data mining to comb PubMed abstracts as well as open access full-text articles and papers published by Elsevier. For example, typing in "alcohol" and "heart disease" brings up a slew of excerpts from studies, highlighting results that failed to draw a link between alcohol consumption and protection against coronary heart disease.
Yu, who published a paper announcing the new technology in BMC Bioinformatics last week, told ScienceInsider that BioNOT is best used by researchers annotating genes to parse the published information on a particular gene's role in a particular disease or condition.