161,979 new RNA viruses have been uncovered thanks to a machine learning tool, which researchers believe will greatly enhance our understanding of the diversity of life on Earth and assist in finding many more yet-to-be-classified viruses.
Artificial intelligence (AI) has uncovered fascinating details about an essential and diverse group of life forms that exist all around us and beneath our feet.
Using a machine learning tool, researchers have identified 161,979 new RNA virus species, a breakthrough that could significantly enhance the mapping of life on Earth and help uncover millions more viruses that have yet to be characterized.
This monumental study, published in Cell, is the largest-ever investigation into virus species discovery and was conducted by an international team of researchers.
“We’ve been granted insight into a largely concealed aspect of life on Earth, highlighting incredible biodiversity,” stated senior author Professor Edwards Holmes from the University of Sydney’s School of Medical Sciences.
“This represents the highest number of new virus species identified in a single research effort, greatly enriching our understanding of the viruses co-existing with us,” Professor Holmes remarked. “The sheer number of newly discovered viruses is astonishing and merely scratches the surface, paving the way for even greater exploration. We anticipate millions more waiting to be found, and this methodology could also be utilized for identifying bacteria and parasites.”
While RNA viruses are often linked to human illnesses, they thrive in extreme environments worldwide and may be critical to global ecosystems. In this investigation, they were located in places like the atmosphere, thermal springs, and hydrothermal vents.
“The presence of such a vast variety of viruses in extreme environments exemplifies their extraordinary diversity and resilience in the toughest conditions, possibly offering insights into the origins of viruses and other basic life forms,” Professor Holmes noted.
HOW THE AI TOOL WORKED
The researchers developed a deep learning algorithm, named LucaProt, which was used to analyze large sets of genetic sequence data, including extensive virus genomes up to 47,250 nucleotides and complex genomic information to uncover over 160,000 viruses.
“Most of these viruses had already been sequenced and were available in public databases, but their significant divergence meant they remained uncharacterized,” explained Professor Holmes. “They were part of what is often called the sequence ‘dark matter’. Our AI approach was able to systematically organize and categorize this information, revealing the significance of this dark matter for the first time.
The AI tool was trained to process this dark matter and identify viruses based on genetic sequences and the secondary structures of the proteins that all RNA viruses utilize for replication.
This method significantly accelerated the discovery of viruses, a process that would traditionally be very time-consuming.
Co-author Professor Mang Shi from Sun Yat-sen University, the study’s leading institution, shared, “Previously, we depended on laborious bioinformatics pipelines for virus discovery, which limited our exploration scope. Now, we have a much more efficient AI-driven model that provides exceptional sensitivity and specificity, allowing us to explore viral diversity deeper than ever. We intend to apply this model to various areas.”
Co-author Dr. Zhao-Rong Li from the Apsara Lab of Alibaba Cloud Intelligence remarked, “LucaProt exemplifies a significant melding of advanced AI technology and virology, showing that AI can effectively facilitate tasks in biological exploration. This synthesis offers valuable insights and motivation for further decoding biological sequences and understanding biological systems from a fresh perspective. We aim to continue our research on AI in virology.”
Professor Holmes affirmed, “The clear next step is to refine our method to uncover even more of this wonderful diversity, and who knows what surprises we might discover next.”
