Carnegie Mellon scientists unveil new tool to understand evolution of multi-domain genes

Results upend current analyses, herald new way to understand, exploit key proteins in cancer

PITTSBURGH—Carnegie Mellon scientists have discovered critical flaws in the standard method used to analyze gene evolution. Standard methods fail when applied to genes that encode multi-domain proteins, an important class of proteins crucial to human health.

Computational biologist Dannie Durand and colleagues have for the first time tackled the dilemma of how to study the ancestry of multi-domain genes.

Correctly identifying gene ancestry is a linchpin of computational genomics. Genes passed down from a common ancestor tend to perform similar functions in the cell. Scientists exploit this similarity to perform tasks such as predicting gene function, mapping human chromosomal regions to corresponding regions in model organisms, and reconstructing the regulatory circuitry that turns genes on and off.

Although computational biologists have developed methods to identify genes that share a common ancestor, current methods often lead to spurious conclusions when applied genes encode multi-domain proteins. Domains are sequence fragments that encode the basic building blocks of protein structure. Evolution makes new genes by mixing and matching domains in novel combinations, much like a child who builds a house, a car and a helicopter from the same LEGO kit by combining LEGO blocks in different ways. This process, called domain shuffling, creates complex proteins that perform specific, critical tasks such as cell communication and binding to other cells. When one of these proteins fails, cancer is often the result. Domain shuffling allows rapid evolution of new proteins, but it also makes it close to impossible for scientists to determine their ancestry.

In a paper published online in Public Library of Science Computational Biology today, Durand’s team presents a novel method to determine whether a pair of similar genes evolved from a common ancestor, or whether they just look similar because the same domain was inserted into both genes. Their method, called “Neighborhood Correlation,” is the first to tackle this problem.

“We needed a completely new approach to determine which multi-domain proteins share a common ancestor, and we are the first group to propose such a method,” Durand said. “Ours is the first approach to define and analyze common ancestry in a traditional vertical way, even when domain shuffling occurs.”