We design enzymes using AI and industry domain knowledge
Our AI model designs new enzymes or enhances existing ones through specific mutations that simultaneously improve multiple properties. Trained on both public and extensive proprietary data, our protein-specific AI models excel in designing FTO and patentable enzymes.
By incorporating human intelligence into protein sequence designs, we simplify downstream analysis and achieve higher data utilization.
We generate new modalities of data to train our models
These AI-generated enzymes are synthesized and tested using our unique ultra-high-throughput one-pot screening platform, which converts diverse biological signals into abundant, quantifiable data under various conditions.
This large, high-quality data is fed back into our AI for continuous learning. Our internally trained AI model achieves a hit rate two orders of magnitude higher than models trained solely on public databases.
Wet-lab innovations will lead the AI revolution in biology
By integrating AI design with advanced screening, we’ve reduced the design-build-test-learn cycle to less than a month, significantly faster than the traditional 3-6 months. Our approach efficiently evaluates enzyme performance across various criteria, effectively addressing the challenge of multi-property optimization.
