AI scans RNA 'dark matter' and uncovers 70k new viruses

Researchers have utilized artificial intelligence (AI) to discover 70,500 previously unknown RNA viruses, many of which inhabit extreme environments such as hydrothermal vents and salt lakes. This breakthrough was achieved through metagenomics, allowing scientists to analyze environmental genomes without culturing individual viruses. The study, published in the journal Cell, highlights the vast unexplored diversity of viruses, with some potentially linked to human diseases. The AI model, named LucaProt, was developed using a transformer architecture similar to that of ChatGPT and was trained to identify viral RNA-dependent RNA polymerase (RdRp) sequences in genomic data. This method revealed approximately 160,000 RNA viruses, with nearly half being new to science. The findings suggest that there are significant pockets of RNA virus biodiversity that remain largely uncharacterized. The researchers emphasize the importance of understanding these viruses, including their potential hosts, to gain insights into their ecological roles and evolutionary history. Future work will focus on predicting the hosts of these newly identified viruses, particularly in relation to archaea, a group of microorganisms for which no RNA viruses have been clearly identified.

- AI has identified 70,500 new RNA viruses, many from extreme environments.

- The study utilized metagenomics to analyze environmental genomes without culturing.

- The AI model LucaProt was trained to recognize viral sequences, revealing significant biodiversity.

- Nearly half of the identified viruses were previously uncharacterized.

- Future research aims to determine the hosts of these new viruses, especially in archaea.