Researchers have utilized artificial intelligence (AI) to identify 70, 500 new viruses that were previously unknown to science, many of which are unusual and unlike any documented species. The identification of these RNA viruses was achieved through metagenomics, a process that allows scientists to sample all the genomes in an environment without needing to culture individual viruses. This method highlights the potential of AI in exploring the 'dark matter' of the RNA virus domain. Viruses are omnipresent microorganisms that can infect animals, plants, and even bacteria; however, only a small number have been identified and documented. Artem Babaian, a computational virologist at the University of Toronto in Canada, states that there is "essentially a bottomless pit" of undiscovered viruses. Some of these viruses may be responsible for diseases in humans, meaning that detailing them could shed light on unexplained illnesses, according to Babaian. Previous research has employed machine learning techniques to discover new viruses within sequencing data. The current study, published in Cell this week, advances this research by examining predicted protein structures. The AI model incorporates a protein-prediction tool named ESMFold, which was created by scientists at Meta (formerly Facebook, based in Menlo Park, California). Another similar AI system, AlphaFold, was developed by researchers at Google DeepMind in London, who recently won the Nobel Prize in Chemistry. Missed Viruses In 2022, Babaian and his team analyzed 5. 7 million genomic samples stored in publicly accessible databases and discovered almost 132, 000 new RNA viruses. Other research teams have also undertaken similar initiatives. However, RNA viruses evolve rapidly, and current identification methods likely overlook many. A standard approach is to search for a portion of the genome that encodes a crucial protein involved in RNA replication known as RNA-dependent RNA polymerase (RdRp).

If the sequence for this protein in a virus differs significantly from known sequences, researchers may fail to recognize it. Shi Mang, an evolutionary biologist at Sun Yat-sen University in Shenzhen, China, and a co-author of the Cell study, along with his colleagues, sought out previously uncharacterized viruses in publicly available genomic datasets. They devised a model called LucaProt, employing the 'transformer' architecture that also supports ChatGPT, and integrated sequencing as well as ESMFold protein-prediction data. They subsequently trained their model to identify viral RdRps and utilized it to locate sequences that encoded these enzymes, which serve as indicators that the sequences belong to a virus, within a vast array of genomic data. Through this method, they identified approximately 160, 000 RNA viruses, including some remarkably lengthy ones from extreme environments like hot springs, salt lakes, and the atmosphere. Nearly half of these viruses were previously unrecorded. Babaian describes discovering "little pockets of RNA virus biodiversity that are truly far off in the boonies of evolutionary space. " “This approach is incredibly promising for broadening the virosphere, ” remarks Jackie Mahar, an evolutionary virologist at the CSIRO Australian Centre for Disease Preparedness in Geelong. By characterizing these viruses, researchers will gain insights into their origins and the evolution of different host interactions. Furthermore, expanding the repository of known viruses facilitates the discovery of additional similar viruses, according to Babaian. “Suddenly, you can see things that you previously overlooked. ” The team was unable to ascertain the hosts of the viruses they discovered, which warrants further investigation, according to Mahar. Researchers are particularly keen to determine whether any of the new viruses infect archaea, an entire lineage of life where no RNA viruses have been convincingly demonstrated to infect. Shi is currently developing a model to predict the hosts of these newly discovered RNA viruses, hoping it will aid researchers in understanding the roles these viruses play within their environmental niches.