Animals and AI help scientists study pandemics

Researchers at the University of Montreal have developed an artificial intelligence model that could help identify and predict areas at risk of emerging viral infections that can jump from animals to humans. By detecting these "hotspots" in advance, they hope to prevent future pandemics like COVID-19. The team of custom software developers spent three years and 10,000 hours of computing to create the algorithm.

Instead of manually analysing data, the algorithm utilises machine learning to assess thousands of mammal species and viruses to identify potential interactions between them. Currently, scientists are only aware of a small fraction of virus-mammal interactions, making it crucial to determine which viruses are likely to infect which mammal species. This information helps establish which interactions are most likely to occur.

To train the algorithm, the researchers used various datasets, including CLOVER, which described thousands of interactions between viruses and mammalian hosts, predominantly focused on wild animals. Other datasets, such as HP3, EID2, and GHMPD2, were also used. The process of training the machine learning model was time-consuming due to the need to clean and verify the data.

The researchers specifically examined 20 viruses that were of concern and had the potential to infect humans. Some unexpected results emerged, such as identifying the Ectromelia virus, linked to mice, as a virus to watch. Further investigation confirmed that cases of this virus affecting humans had indeed been reported.

The model also revealed two regions of interest: the Amazon basin, known for unique virus-host interactions and the likelihood of new interactions, and Sub-Saharan Africa, where the algorithm identified new hosts likely to carry zoonotic viruses. This information can guide scientists in targeting their viral and vaccine research efforts.

Zoonotic diseases, whether bacterial, parasitic, or viral, are expected to become more prevalent as human and animal habitats increasingly overlap. The researchers hope that their model can not only guide future research but also contribute to real-time surveillance. The next step is to enhance the AI model by incorporating additional microbiological, immunological, and ecological mechanisms to obtain a comprehensive understanding of the global viruses. Hopefully, they could get AI efficiently working just as it did in matching drugs for patients.

In 2016, the United Nations Environment Programme (UNEP) raised concerns about zoonotic diseases, and the World Health Organization reports that zoonotic diseases cause about one billion cases and millions of deaths annually. Shockingly, 75% of the 30 new human viruses discovered in the last 30 years originated in animals.

The algorithm developed by the University of Montreal could mitigate the impact of zoonotic diseases and potentially assist in preventing future pandemics by guiding targeted research and surveillance efforts.