Researchers at the University of Oregon have developed an artificial intelligence tool that can read genetic code the way large language models like ChatGPT read text.\u00a0Scanning the genome for biological mutation patterns, the computer model traces pairs of genes back in time to their last common ancestor.\u00a0<\/p>\n
It’s\u00a0the first language model designed for population genetics, said\u00a0Andrew Kern, a computational biologist in the UO\u00a0College of Arts and Sciences. As described in a paper published April 10 in the\u00a0Proceedings of the National Academy of Sciences, the AI tool offers scientists a fast and flexible alternative to classical methods for reconstructing evolutionary history.\u00a0<\/p>\n
In practice, it\u00a0can help\u00a0researchers like Kern\u00a0understand when disease-resistance genes\u00a0emerged\u00a0in a population, for example, or when\u00a0species\u00a0evolved key traits.\u00a0<\/p>\n
“Advances in generative AI and the\u00a0architectures\u00a0behind them are potentially useful to a number of fields outside a chatbot,” said Kern, an Evergreen professor of biology. “We’re borrowing strengths from the world of AI and applying them in this different context that’s largely been untapped.”\u00a0<\/p>\n
Genomes are often compared to a written language, with combinations of DNA’s four-letter alphabet – A, T, C and G – forming the basis for genes and chromosomes. Kern and his lab are most interested in\u00a0what’s\u00a0misspelled, which scientists call mutations: changes in DNA sequences, like swapped or missing letters, that accumulate over time as part of evolution.\u00a0<\/p>\n
Often harmless, mutations can be passed down from generation to generation, leaving a trail of breadcrumbs for tracing ancestral relationships.\u00a0<\/p>\n
Traditional methods based on math and statistics are the gold standard for translating mutations into ancestry.\u00a0They’re\u00a0difficult to beat in most cases, said Kevin Korfmann, lead author of the study and former postdoctoral researcher at the UO. But\u00a0those\u00a0classical probabilistic approaches can be slow and struggle with large or incomplete genomic datasets, he\u00a0added.\u00a0<\/p>\n
So, the researchers looked\u00a0to\u00a0AI to efficiently interpret the language of life by\u00a0modifying\u00a0a GPT-2 model, the older machine learning architecture behind ChatGPT. But instead of being trained on large volumes of English text, the language model was trained on simulations of genetic evolution across\u00a0different species\u00a0– including bacteria, rodents,\u00a0mosquitoes\u00a0and primates – to learn and recognize mutation<\/a> patterns.\u00a0<\/p>\n “We can’t repeat evolution, so one of the key workflows we have is developing simulations,” Korfmann said. “The simulations\u00a0mimic\u00a0evolutionary processes, and then we use the outcomes as training data for our deep learning models.”\u00a0<\/p>\n In general, stretches of DNA with many mutations likely trace back to a distant common ancestor, whereas those with few mutations are likely to share a more recent ancestor.\u00a0This helps explain why chimpanzees are considered humans’\u00a0closest living relatives, with similar DNA, while sea sponges are the most distant, having diverged\u00a0genetically\u00a0more than 700 million years ago.\u00a0<\/p>\n Based on those mutation patterns and other biological principles, the AI model can predict when gene pairs last shared a common ancestor, known as the “coalescence time.”\u00a0<\/p>\n In tests, the tool performed as well as\u00a0state-of-the-art\u00a0statistical methods,\u00a0which was\u00a0surprising\u00a0to\u00a0the research team.\u00a0<\/p>\n “You never really know what’s going to work when you’re essentially borrowing techniques from a totally different world and applying them to a new problem,” Kern said. “But this was a case where things worked really well.”\u00a0<\/p>\n The computer model\u00a0was\u00a0also dramatically faster.\u00a0While\u00a0traditional methods can take hours or even days to decode\u00a0a single mosquito chromosome,\u00a0the\u00a0new approach\u00a0can do it in minutes.\u00a0That\u00a0efficiency is especially\u00a0beneficial for\u00a0scientists\u00a0handling\u00a0large amounts\u00a0of genetic sequence data.\u00a0<\/p>\n “Compared to classical inferential approaches, the AI tool doesn’t have to reason\u00a0about\u00a0every mutation individually,” Korfmann said. “It just reads the patterns because all of the expensive statistical work was done up front, during training, which sidesteps the bottleneck.”\u00a0<\/p>\n The model’s simulation-based training also enables scientists to use DNA datasets that are incomplete or missing genetic code – an issue Kern\u00a0frequently\u00a0faces when working with mosquito genetic databases for\u00a0his research on malaria transmission.\u00a0<\/p>\n That versatility comes at a crucial moment for malaria control, Kern said. For decades, insecticides have been a cornerstone for\u00a0the\u00a0prevention of malaria-spreading mosquitoes. But evolution, as Kern puts it, “did its thing.”\u00a0<\/p>\n “Insecticide resistance is being observed in all of these mosquito populations today,” he said.\u00a0“A major challenge in preventing the spread of malaria has been understanding the evolution of insecticide resistance. Now, we can go in with our AI model, ask how long ago these resistance genes arose in the population, and learn about the evolutionary history of this critical carrier of malaria.”\u00a0<\/p>\n Looking ahead, Kern and Korfmann\u00a0aim to advance the biological model\u00a0beyond tracing shared ancestry between\u00a0two\u00a0lineages\u00a0towards\u00a0reconstructing full\u00a0genealogical\u00a0trees across multiple\u00a0lineages.\u00a0Some traditional methods can already do this,\u00a0but Kern said\u00a0they’d\u00a0like to\u00a0chase\u00a0that goal\u00a0from\u00a0a\u00a0machine-learning angle.\u00a0<\/p>\n “There’s so much going on in the machine learning field that we haven’t applied yet in our field,” Korfmann said. “There’s tons of translational work to do to get these novel algorithms working in biology.”\u00a0<\/p>\n Source:<\/p>\n Journal reference:<\/p>\n