This article has been updated to clarify the affiliations of the researchers.
Researchers at the Virginia Polytechnic Institute have developed a computational method to study the mechanisms of infectious disease that could help identify novel therapies against a range of pathogens, according to the Virginia Tech team.
In the study, which was published last week in PLoS Pathogens
, the researchers from the Virginia Bioinformatics Institute and the department of computer science at Virginia Tech integrated experimentally verified human-pathogen protein-protein interactions from seven public databases into a single intraspecies network covering around 1,200 human proteins.
This integrative approach has uncovered a characteristic of pathogenic mechanisms that was not evident from any of the seven public databases alone: Pathogens appear to have evolved to interact with human protein “hubs,” which are involved in many interactions, and “bottlenecks,” which are central to many pathways.
One key aspect of the study, according to senior author Bruno Sobral, is that “because we took a view of the forest rather than the trees — in other words, we looked at many pathogens interacting with human hosts — there seem to be, based on these computational predictions, certain types of host proteins that are targeted over and over and over again by many pathogens.”
Sobral noted that this finding is timely in light of the National Institute of Allergy and Infectious Diseases’ recent mandate to move away from the so-called “one bug, one drug” strategy in favor of medical countermeasures that are effective against a variety of pathogens and toxins, as outlined in its 2007 Strategic Plan for Biodefense Research
“Because we’ve taken this view of the forest, it’s exactly this kind of approach that can help us determine whether, if a protein gets hit over and over by 10 different viruses, it is going to therefore be a good target for us to think about building one countermeasure that would work against all 10 different viruses.” Sobral said. “That’s never been done before.”
In the study, Sobral and Matthew Dyer of VBI and T.M. Murali of the Virginia Tech computer science department integrated 10,477 known interactions between human proteins and proteins from 190 pathogens from 54 taxonomically related groups — 35 viral, 17 bacterial, and two protozoan.
“This looks like a very real demonstration of properties that people have observed about interspecies protein interaction networks by simulation.”
Gene Ontology analysis of highly targeted human proteins found that pathogens tend to target proteins involved in processes like cell-cycle regulation, nuclear transport, and immune response.
In addition to the implications of the study for therapeutic development, the findings may help settle a longstanding debate over the fundamental characteristics of protein-interaction networks, according to Murali. Other groups that have studied these networks within a single species, such as yeast, have observed that they exhibit power-law properties — that is, there are very few highly connected proteins and very many minimally connected proteins.
However, Murali said, “there is a lot of controversy about whether the power-law behavior is actually an intrinsic feature of protein-interaction networks or whether it’s an artifact of experimental biases.”
Murali said that the VBI study serves as “an actual biological demonstration” of the power-law property. “We have these external agents, the so-called pathogens, which seem to have evolved the ability to preferentially interact with hubs and bottlenecks. So in my mind, this looks like a very real demonstration of properties that people have observed about interspecies protein-interaction networks by simulation.”
The VBI researchers caution that their results “should be interpreted with caution since no single pathogen may target all the proteins and PPIs we analyze.”
They propose, however, that “piecing together targeted human proteins across multiple pathogens has the potential to provide insights into common molecular mechanisms of infection and proliferation used by different pathogens.”
Wanted: More Data
The VBI team is currently assembling new data to carry out further work. One drawback they experienced when collecting data sets for the study was a relative dearth of bacterial human-pathogen PPIs; 98.3 percent of the interactions they used are from viral systems, and the majority of those are from human-HIV studies.
Dyer said that he and his colleagues are working with Myriad Genetics to gain access to human-bacteria interaction data for three pathogens: Yersinia pestis, responsible for bubonic plague; Francisella tularensis, which causes tularemia, or rabbit fever; and Bacillus anthracis, the infectious agent behind anthrax.
The VBI team has only studied one of those data sets so far, and Dyer said that initial results indicate that the bacterial proteins “tend to target human hubs and bottlenecks at a statistically significant level,” just as the viral pathogens do.
Sobral said that the researchers are prioritizing potential therapeutic targets in the context of several large-scale genomics and proteomics experiments that NIAID is currently funding under its biodefense initiative.
“A first validation step is to look at those high-throughout data, the proteomics and transcriptomics data sets … to see whether the predictions we’re making match, and how they match, and how they don’t match when they don’t — the actual things that are being observed in the experiments,” he said.
Sobral said that one goal of the project is to create a “computational engine” that will be able to predict targets for vaccines or therapies entirely in silico. “To get there, of course, we have to go through all these iterations where we first build the computational engine, then, through our partnerships, we validate that engine and improve that engine based on the results that we’ve observed, and then the next step is, based on a small subset of those results, people would use them as potential targets and try to develop a drug or a vaccine.”
The researchers stressed that the intraspecies interaction map is just a first step toward identifying potential therapeutic targets. Since so many of the proteins that pathogens target are also crucial to the host’s survival, countermeasure development isn’t as simple as identifying “hub” proteins.
“One thing you have to consider is gene-essentiality data,” Dyer said. “You don’t want to be targeting human proteins that are essential to the cell, so that if you target those then the cell dies. So there are other sorts of data that need to come into play here when you’re trying to find good, drugable targets.”