AI model trained on 9.3T base pairs can now design novel genes (www.dongascience.com)

Scientists have unveiled 'Evo 2,' a groundbreaking AI model trained on an unprecedented 9.3 trillion base pairs from the genomes of over 128,000 species. Developed by a collaboration of researchers from the Arc Institute, NVIDIA, Stanford University, UC Berkeley, and UCSF, Evo 2 excels in identifying genetic mutations linked to diseases such as breast cancer and designing artificial genomes at the bacterial level. This model can process genomic sequences up to one million base pairs at a time, significantly enhancing our understanding of gene relationships and paving the way for advancements in genetic analysis. Evo 2's significance lies in its potential to revolutionize genetic research and healthcare. Its ability to predict disease risks and evaluate genetic variations can accelerate drug development and save valuable time compared to traditional experimental methods. The model has already demonstrated its utility by accurately identifying pathogenic mutations in the BRCA1 gene with over 90% accuracy and aiding in the design of bacteriophages, which may help combat antibiotic-resistant bacteria. However, the journey towards creating fully functional synthetic life remains challenging, with experts cautioning that even minor genetic errors can render artificial genomes nonviable. Evo 2's foundational architecture opens avenues for further specific AI models, promising a myriad of applications in biological science and beyond.