Researchers use AI to study deadly hemorrhagic fever viruses
As part of a $3.25 million Department of Defense grant through the Defense Threat Reduction Agency, researchers at the Virginia-Maryland College of Veterinary Medicine, the University of Montreal, and SRI International are using artificial intelligence (AI) to study hemorrhagic fever viruses with the ultimate goal of identifying targets for therapeutics.
The researchers are starting their project with Rift Valley fever virus (RVFV), a hemorrhagic fever virus with no FDA-approved treatment or vaccine.
Spread by mosquitoes, RVFV is found primarily in sub-Saharan Africa. It most commonly affects domesticated ruminants like sheep, goats, and cattle, though RVFV can also infect humans.
"It's a really important virus: it causes significant disease in livestock as well as in humans, and it has a huge economic impact as well as a public health impact. For humans, the disease is milder, but it can cause severe liver disease, ocular issues, and encephalitis as well,” said Kylene Kehn-Hall, professor of virology and the principal investigator on the project.
Kehn-Hall is collaborating with James Omichinski, professor at the University of Montreal, and with Biocomplexity Sciences program director Paul O’Maille and senior computer scientist Andrew Silberfarb of SRI International, an independent nonprofit research and development institute. Additionally, doctoral students are working on the project — Kaylee Petraccione, a PhD student in Kehn-Hall’s lab, co-authored a paper that was published in PLoS Pathogens in March.
The majority of humans infected with RVFV suffer from mild, flu-like symptoms, but about 10% of infected humans experience extreme symptoms, including hemorrhagic fever.
About 30% of adult ruminants infected with RVFV die, and for young ruminants, the mortality rate is nearly 100%. One major sign of a Rift Valley fever outbreak is en masse abortions in ruminants: almost 100% of infected pregnant ruminants experience spontaneous abortions. This virus is not only devastating to ruminant populations but also for the people who rely on them for their food and livelihood.
The Department of Defense is invested in hemorrhagic fever virus research for several reasons.
“The viruses we're working on are considered biodefense pathogens — they could be used for nefarious purposes. They're also a threat to military personnel who go into different environments and may be exposed to these pathogens,” Kehn-Hall explained.
For this project, the researchers are studying short linear motifs (SLiMs), protein sequences that facilitate interactions between proteins.
"We're looking for a 'barcode' within a viral protein — we’re looking to identify the sequence, and, using AI, predict whether it's going to interact with another key protein in our body. We want to know if the viral proteins will interact with particular host proteins and if that interaction is important for the virus to replicate,” said Kehn-Hall.
Through studying SLiMs, researchers can identify specific interactions between host cell proteins and viral proteins, which can act as targets for future therapeutics. Drugs can be developed to block the viral protein-host protein interactions that cause the virus to spread.
The researchers are using a computational program to model how proteins make a 3D structure. AI uses that information to predict the likelihood of proteins binding to one another, identifying potential interactions.
"Any one interaction could take a year to a couple years to study, but with AI, we can predict hundreds of interactions within only a few months,” said Kehn-Hall. Not all of the AI’s predictions are correct, but once those interactions have been flagged by AI, researchers can test the predictions through classical molecular biology and virology experiments.
The more the researchers work with the AI and give it more information through wet lab work, the more accurate its predictions will become.
In the paper recently published in PLoS Pathogens, Petraccione and collaborators focus on SLiMs connected to autophagy. Infection with a virus like RVFV triggers autophagy, a process where cellular material is broken down and reused. Through autophagy, the body’s cells can break down viral components as a defense mechanism — however, viruses have developed their own ways to elude this antiviral immune response. This paper shines light on the mechanism RVFV uses to suppress autophagy and spread across the cell.
Over the coming years, the researchers will turn their attention to other hemorrhagic fever viruses. By the end of this process, they will have amassed a large library of information about the viruses and will have trained an AI to predict protein interactions — so when the next hemorrhagic fever virus emerges, we will be better prepared to tackle the threat.
