![[Reel Two]](logo.gif){kind=logo}
In addition to our products, Reel Two has a number of ongoing
research projects. Some were started as technology demonstrators, while
others represent functionalities that will eventually be productized,
or incorporated into other products and solutions.
SureGene
SureGene applied Reel Two's machine learning expertise and Classification System technology to the problem of resolving ambiguous gene names and synonyms. The goal was to identify PubMed abstracts that were genuinely about the query gene, and filter out papers referring to non-relevant synonyms or non-related genes. Scientists working with Reel Two were "impressed by its ability to narrow the range of papers to [desired] targets and weed out the ones of limited relevance, and by the accuracy it delivers through the use of all synonyms for the genes."SureGene operates by automatically assigning gene names to their
Entrez-Gene
ID (a.k.a. LocusLink ID) in PubMed abstracts. Classification models for
each gene are used to analyze the
context of the articles, rather than just searching for keywords. As a
result, SureGene searches are not hindered by the presence (or absence)
of gene synonyms,
incorrect gene name references or non-genetic synonyms. To date
SureGene covers 8,300 human genes,
which includes most genes well characterized in the literature. Reel
Two plans to use the classification schema and models developed for
SureGene as part of enhancements for other Reel Two products.
For those interested in seeing how effective SureGene is at
filtering documents by the relevant gene name, an online demo of
SureGene is available.
Gene Ontology Knowledge Discovery System
Reel Two's Gene Ontology Knowledge Discovery System (GO KDS) was designed to show how Reel Two's Classification System could be applied to a large taxonomy and a corpus of millions of documents. The result was a web-based search service enabling researchers to retrieve MEDLINE abstracts according to Gene Ontology terms. GO KDS classifies the full collection of 12 million MedLine® entries - the most comprehensive collection of scientific abstracts annotated by GO terms.
Documents are listed by their relevance to the queried GO terms. In addition each document displays a list of other GO terms by which it has been classified. This allows users to see relationship between an abstract and GO terms from all three branches of the Gene Ontology. Users can also use GO KDS to classify their own documents by GO term.
Powered by Reel Two's patent-pending Classification System, GO KDS has increased the number of GO-annotated publications by 400 times. More than 4000 GO terms have been modeled by GO KDS, displaying the ability of Classification System to handle thousands of categories, hundreds of thousands of training examples and millions of documents.
To access the public version of GO KDS, register at http://www.go-kds.com. For information on full-featured, interactive versions of GO KDS or other text mining portal solutions, please contact sales@reeltwo.com.