Many new diseases that affect us humans find their origin in viruses derived from the world of animals. This is probably also the case for the infamous SARS-CoV-2 virus, colloquially known as the coronavirus.
Unfortunately, identifying which animal viruses are at high risk of jumping over to humans has proven very challenging and time-consuming. Scientists have determined that machine learning can be the key here.
AI may help us stop the next pandemic before it even happens.
Image Credit: Gorodenkoff via Shutterstock / HDR tune by Universal-Sci
Identifying possible zoonotic viruses
An infectious disease triggered by a virus or parasite that has jumped from an animal to a human is called a zoonotic disease or zoonosis.
Recognizing zoonotic viruses before they arise forms a big issue because just a small percentage of the approximate 1.67 million animal viruses are actually capable of infecting humans.
When researchers discover a new virus among animals, it is very difficult to quickly assess whether it has the potential to make the transition from animals to humans and therefore also to determine whether this virus merits the investment of further research.
But a new study, published in the science journal PLOS Biology, may make the lives of researchers chasing zoonoses a little easier. In the study, scientists present a method based on the genome of a virus (often the only thing we know about newly discovered or poorly characterized viruses) to determine whether it is able to jump from animals to humans.
The concept of zoonosis - (Image Credit: Crystal Eye Studio via Shutterstock)
Using machine learning to identify dangerous viruses
The science team initially gathered a dataset of 861 virus species from 36 different families in order to construct machine learning models utilizing viral genome sequences.
They created machine learning algorithms that used patterns in viral genomes to calculate the likelihood of human infection. The best-performing algorithm was then used to look for trends in the projected zoonotic potential of other viral genomes from a variety of species.
The team discovered that viral genomes may have generalizable features that are independent of virus taxonomic connections and may adapt the viruses for life in conditions it has yet to encounter (IE to infect humans).
Eventually, they succeed in creating machine learning models suitable to distinguish potential zoonotic diseases using viral genomes. However, these models have shortcomings, as computer models are only a preparatory step of distinguishing zoonotic viruses.
Viruses flagged by the models will need further lab testing before confirmation. Mainly because deciding to do additional research on a virus is an impactful decision tied to large investments. One cannot simply make a decision without careful deliberation.
On top of that, although these models can help forecast whether a particular virus may be capable of infecting people, the capacity to infect is just a single part of broader 'zoonotic risk.' Other relevant factors are the virus's ability to transmit between humans, its virulence in humans, and the ecological circumstances at the time of human exposure.
Image Credit: Corona Borealis Studio via Shutterstock / HDR tune by Universal-Sci
Preventing the next pandemic
According to the researchers, these results demonstrate that the zoonotic potential of viruses can be inferred to a surprisingly large extend from their genome sequence.
Simone Babayne (lecturer at the University Institute of Biodiversity Animal Health & Comparative Medicine and one of the authors) adds that as more new viruses get described, the more effective their models will get.
AI techniques will become better and better at identifying high-risk viruses giving us more time to research and monitor them. In time scientists may even be able to develop a preemptive vaccine before an outbreak arises among the larger population. All in all, these are very hopeful developments.
Sources and further reading:
If you enjoy our selection of content, consider subscribing to our newsletter
FEATURED ARTICLES:
