Gene Ontology Database

National Genome Research Institute (NHGRI, Washington, DC, USA).

Biomind LLC (Rockville, MD, USA; www.go.biomind.com) is now applying advanced machine learning algorithms to guess the appropriate category for classifying hundreds of previously unclassified genes. The Biomind software makes its guesses by combining information about gene sequences with data mined from a large body of microarray data, incorporating data from a variety of different experiments. The results of this analysis are available on the company's website.

The basic concept is that if a gene behaves and looks similar to the genes in some Gene Ontology category, such as "DNA repair” or "glycolysis,” then perhaps the gene belongs in that category. However, it is not always that simple. "Getting from this basic concept to a working classification system with high accuracy involves a lot of computer science,” explained Dr. Ben Goertzel, CEO of Biomind. Statistical testing suggests that the system's prediction will be roughly 80% accurate, and this accuracy may be improved in the near future by utilizing a broader collection of microarray data.

According to Dr. Goertzel, this information on gene function "should be interesting to anyone doing genomics research. It suggests possible functions for a whole lot of genes that were previously not understood at all, some that were never mentioned in the research literature, because the biologists haven't gotten around to studying them yet. For instance, there are dozens of genes that our software says are associated with programmed cell death--a key cause of aging--and that weren't previously known to have anything to do with anything.”





Related Links:
Gene Ontology
Biomind