Functional gene clustering via gene annotation sentences, MeSH and GO keywords from biomedical literature



Jeyakumar Natarajan1, * and Jawahar Ganapathy1



1Centre of Excellence in Bioinformatics, School of Biotechnology, Madurai Kamaraj University, Madurai 625021, India


Email; * Corresponding author


Article Type




received December 14, 2007; accepted December 30, 2007; published online December 30, 2007



Gene function annotation remains a key challenge in modern biology. This is especially true for high-throughput techniques such as gene expression experiments. Vital information about genes is available electronically from biomedical literature in the form of full texts and abstracts. In addition, various publicly available databases (such as GenBank, Gene Ontology and Entrez) provide access to gene-related information at different levels of biological organization, granularity and data format. This information is being used to assess and interpret the results from high-throughput experiments. To improve keyword extraction for annotational clustering and other types of analyses, we have developed a novel text mining approach, which is based on keywords identified at the level of gene annotation sentences (in particular sentences characterizing biological function) instead of entire abstracts. Further, to improve the expressiveness and usefulness of gene annotation terms, we investigated the combination of sentence-level keywords with terms from the Medical Subject Headings (MeSH) and Gene Ontology (GO) resources. We find that sentence-level keywords combined with MeSH terms outperforms the typical ‘baseline’ set-up (term frequencies at the level of abstracts) by a significant margin, whereas the addition of GO terms improves matters only marginally. We validated our approach on the basis of a manually annotated corpus of 200 abstracts generated on the basis of 2 cancer categories and 10 genes per category. We applied the method in the context of three sets of differentially expressed genes obtained from pediatric brain tumor samples. This analysis suggests novel interpretations of discovered gene expression patterns.



text mining; functional clustering; microarray data analysis


Natarajan & Ganapathy, Bioinformation 2(5): 185-193 (2007)


Edited by

T. W. Tan & S. Ranganathan






Biomedical Informatics


This is an Open Access article which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. This is distributed under the terms of the Creative Commons Attribution License.