Models in DNA Reveal Hundreds of Unknown Protein Matching

Sequencing of a genome becomes cheaper, but it remains difficult to make sense of the data obtained. The researchers have now found a new way to extract useful information from sequenced DNA.

By listing the subtle evolutionary signatures shared between pairs of genes in bacteria, the team was able to uncover hundreds of previously unknown protein interactions. This method is now applied to the human genome and could provide a better understanding of the interaction of human proteins.

The project is a collaboration between scientists from the Faculty of Medicine at the University of Washington and Harvard University. Their report appears in the July 11 issue of Science.

"Protein-protein interactions are fundamental to biological function, and it is remarkable to be able to predict them massively using the large amounts of genomic sequence data generated in recent years," said David Baker, senior author, professor of biochemistry. at the university. from the Washington School of Medicine.

The cells are full of proteins, many of which must physically interact to function. This may mean coming together to copy DNA or to form long fibers like those found in the muscle. In many cases, however, scientists still do not know which proteins interact. Finding new couples can be slow, laborious and expensive.

In search of a better solution, a team of four computer biologists studied a phenomenon called coevolution, in which changes in one gene are associated with changes in another. This may indicate that two genes are significantly related.

For example, if one gene mutates to produce a modified form protein, a second one can evolve to produce a protein of complementary shape to the first, thus preserving the interaction capacity of the two proteins.

In recent years, researchers have found evidence of some of these subtle molecular interactions in the DNA of an organism.

"Coevolution has been helpful in understanding how specific proteins interact, but we can now use it as a discovery tool," said lead author Qian Cong, a postdoctoral researcher at UW School of Medicine. .

The research team compared more than 4,000 E genes. Coli to DNA sequences from over 40,000 other bacterial genomes. This important stock of genetic information allowed researchers to use a tailored statistical model to assess the co-evolution between each E gene. Coli.

After several rounds of analysis, it was found that 1,618 pairs had the strongest evidence of co-evolution. By comparing their results to a small set of already characterized protein-protein interactions, the researchers obtained a considerably higher accuracy than previous experimental screening methods.

Of the newly discovered interactions, some suggest new biological knowledge. One of these, an interaction between a toxin protein and its antitoxin, could help, according to the researchers, to explain why some E. coli dominate their microbial niche. Another recently discovered pairing suggests that a protein called PstB, known to play a role in metabolism, could also help coordinate protein synthesis and mineral transport.

"In biology, it's rare that a software tool produces enough promising predictions to be tested, but that's exactly what's happening here," Cong said. There are literally hundreds of follow-up experiments that could be performed in laboratories around the world. "

The team also explored the genome of Mycobacterium tuberculosis, a pathogenic bacterium distantly related to E. coli. They identified 911 protein-protein interactions with high confidence. 95% of them had never been described before. Seventy involve proteins that can contribute to the virulence of M. tuberculosis, the researchers report. These findings could open new avenues for developing drugs against the deadly pathogen.

"We will apply this tool to more pathogens and the human genome," Cong said. "Our success will depend on the work that other scientists have put in place to annotate which parts of the genome are genes and which parts are different."