New METAGENE-1 Model Released by USC to Enhance Pathogen Detection with 7 Billion Parameters Trained on 1.5 Trillion DNA/RNA Base Pairs

Jan 6, 2025, 05:08 PM

A new state-of-the-art metagenomic foundation model, METAGENE-1, has been released in collaboration with researchers from the University of Southern California (USC). This model, which features 7 billion parameters, was trained on over 1.5 trillion DNA and RNA base pairs collected from wastewater samples. The development aims to enhance planetary-scale pathogen detection and mitigate the risks of pandemics in an era characterized by rapid biological advancements. Initial benchmarking indicates that advanced models like GPT-4o with web access outperform competitors such as Gemini, Claude, BioloGPT, and BioMedLM in various tasks related to genomic knowledge.