BGI-Research and Zhejiang Lab unveil Genos, a powerful AI model trained on 636 genomes to identify disease-causing mutations with high accuracy.
SHENZHEN/HANGZHOU: In a breakthrough for global genomic science, Chinese tech innovators BGI-Research and Zhejiang Lab have jointly introduced Genos, a new artificial intelligence model capable of identifying disease-causing gene mutations with unmatched precision.
Genos is being celebrated as the world’s first deployable genomic foundation AI model with 10 billion parameters, trained specifically to address the limitations of previous models that relied on narrow genetic datasets. Unlike most models trained on one or two reference genomes, Genos is built on 636 complete human genomes from diverse global populations, representing a major leap in both scale and accuracy.
Genos can analyse up to one million DNA base pairs at a time, achieving single-base resolution, which enables researchers to pinpoint harmful mutations with striking clarity. In testing scenarios, the model reached 92% accuracy in identifying pathogenic mutations — and that figure climbed to 98.3% when combined with a scientific foundation model.
With the human genome composed of around 3 billion base pairs, understanding the role of individual bases in health and disease remains a major challenge. Genos aims to bridge this gap by offering a tool that accelerates the interpretation of genetic data and supports personalised medicine research.
The model is also expected to advance rare disease research, cancer diagnostics, and other precision health applications. Its capacity to handle diverse genomic data offers promise in identifying genetic risk factors across multiple ethnicities — a common shortcoming in earlier models.
According to BGI-Research and Zhejiang Lab, the deployment of Genos marks a significant moment in integrating AI with life sciences, potentially setting a new standard in biomedical innovation.
