Trait-to-gene: A computational method for predicting the function of uncharacterized genes

Mitchell Levesque, Dennis Shasha, Wook Kim, Michael G. Surette, Philip N. Benfey

Research output: Contribution to journalArticlepeer-review


The function of unknown genes is often inferred from comparisons to well-characterized homologs. In this paper, we show that, even if all of the homologs of a gene are unannotated, its function may be deduced through phylogenetic profiling. We have designed a series of algorithms that make functional predictions of genes based on orthology and set theory, but our approach to predicting gene function requires no previous knowledge of homolog function. With this technique, we successfully identified 94% of the clusters of orthologous groups that are known to be involved in flagella development or function. As a test, we removed the function of three putative flagellar genes that had been previously uncharacterized in Bacillus subtilis. We observed a motility phenotype for two of these three genes. Thus, these algorithms allow for high-throughput functional prediction of genes beyond that provided by simple orthology-based annotation endeavors.

Original languageEnglish (US)
Pages (from-to)129-133
Number of pages5
JournalCurrent Biology
Issue number2
StatePublished - Jan 21 2003

ASJC Scopus subject areas

  • General Biochemistry, Genetics and Molecular Biology
  • General Agricultural and Biological Sciences


Dive into the research topics of 'Trait-to-gene: A computational method for predicting the function of uncharacterized genes'. Together they form a unique fingerprint.

Cite this