Identifying functional relationships among human genes by systematic analysis of biological literature
1 Life Science Informatics Unit, Functional Genomics Area, Novartis Pharmaceuticals, Summit, NJ 07901, USA
2 Division of Molecular Genetics, Department of Pediatrics, Columbia University, New York, NY 10032, USA
BMC Bioinformatics 2002, 3:16 doi:10.1186/1471-2105-3-16Published: 7 June 2002
The availability of biomedical literature in electronic format has made it possible to implement automatic text processing methods to expose implicit relationships among different documents, and more importantly, the functional relationships among the molecules and processes that these documents describe.
A computational strategy that identifies functionally related human genes by detecting the implicit relationships among the publications cited under each gene in the Online Mendelian Inheritance in Man (OMIM) was implemented. The implementation was based on a substantially modified version of the kernel document method. The improvements include assigning a calculated weight for a document to indicate its importance in establishing the relationship between two documents, and using multiple kernel documents to reflect the multiple functions of the same gene. An example of using this strategy to identify genes related to the apoptosis pathway in human was given.
The results showed that this method can indeed produce meaningful results when applied to human genes.