Bioinformatic reconstruction of global networks provides shortcuts to protein functions

Researchers at the Stockholm Bioinformatics Center in Sweden have developed bioinformatic methods for reconstructing global networks of proteins and genes that interact with each other functionally. Eight different types of large-scale genomic, proteomic, and functional genetic data have been combined in the largest reconstructions ever of networks in nine different species. The work is published in the June issue of the journal Genome Research, and the networks are available to researchers and others via a Web-based database.

The database, developed by Professor Erik Sonnhammer's research team, is called FunCoup, which refers to the functional link between genes/proteins. This link can vary in nature - either direct physical contact, members of a complex, or members of a certain biological process. Mapping biological processes in the form of networks can be seen as the next step following the identification of all genes and the basic classification of their biochemical functions. Each gene and protein exercises its function through interactions with other molecules, usually other proteins or genes. These are the interactions that FunCoup can predict.

"The advantage of combining different types of data is that they complement each other. All forms of experiments have specific weaknesses, but by using a combination strategy, we can obtain an overall picture with better coverage," says Erik Sonnhammer.

Even though FunCoup is based on data that are publically available, in practice it is not possible for an individual researcher to process the data the way SBC does.

"For instance, we calculate the correlation in protein expression across hundreds of experiments between all pairs of proteins in a species. There are some 200 million pairs in human alone. Huge amounts of correlations and other indicators are analyzed statistically in order to translate a raw signal into a unit of functional coupling. A unique feature of FunCoup is that the type of link is predicted, for example, that 'member of complex' may be more probable than 'direct physical contact.' What's more, the networks in several species are analyzed simultaneously in order to determine whether the process is captured in model organisms," says Erik Sonnhammer.

The article in the journal Genome Research shows, for instance, how FunCoup predicts new genes in Alzheimer's, Parkinson's, and cancer.

"We hope that many scientists will discover additional relevant genes in FunCoup for their research. Biomedical researchers are often familiar with only a small number of genes that are important in a disease, and they put all their resources into these. They should more often broaden their perspectives to include other genes that are close by in the network," says Erik Sonnhammer.

Source: Swedish Research Council