Cancer Proteins Show Off Their Networking Skills

New research suggests that cancer proteins, like influential people, have the most connections. These results, from an extensive study of how human proteins interact with one another, could help explain why cancer wreaks such havoc in cells.

“We haven’t gotten to the bottom of what the increased connectivity really means, but perhaps highly connected proteins, once mutated, are more likely to cause disease,” says co-author Paul Bates, PhD, who heads the Biomolecular Modelling Laboratory at the Cancer Research UK London Research Institute. The research was published in the September 15, 2006, issue of Bioinformatics.

Bates and his graduate student Pall Jonsson built a model of the human proteome that contained more than 108,000 interactions using experimental information on proteins in other species, including yeast and worms. The typical protein can connect with a limited number of other proteins. The researchers scored the data and linked proteins known to interact, leading to a protein-protein interaction network, also known as the “interactome.” Then, using information from a 2004 census of 346 human genes known to mutate in cancer, Bates and Jonsson mapped 509 human cancer proteins onto their network.

The average cancer protein was linked to 23 other proteins in the network, more than twice as many as the typical protein. By analyzing protein clusters, the researchers also found the proteins from cancer genes tend to occupy intersections between protein communities that govern crucial functions such as regulating cell growth and death. This makes sense, says Bates, as proteins that are changed by cancerous mutations tend to disrupt many cellular functions. Bates adds that understanding the network properties around these proteins could help researchers identify drug targets.

Shinichiro Wachi, a doctoral candidate at the University of California, Davis, says the results are hard to interpret. The mapped cancer proteins were based on genetic information rather than experimental data on how they interact, he says. “A gene may perform a function in the cell, but the mutation could either reduce the function or result in higher activity. … It could go either way,” says Wachi, who has studied the network properties of proteins in lung cancer tissues. Wachi also cautions that because the list of cancer genes is changing dramatically, the researchers may soon need to re-examine their model.

Bates would like to do further analysis. “We’ve only got 108,000 interactions. There’s likely to be more than that—400,000, maybe 700,000,” he says. “We want to increase the map and validate it further.”