Translate ideas into impact™
Explore current openings and get ready to prove that impossible isn’t.Unlimit with us™
We train our Denovium™ Engine AI models on high-quality, proprietary datasets generated through our synthetic biology platform. We test the AI predictions in the lab and iterate to refine the models, cycling our technology development toward better and more comprehensive capabilities. With each cycle, our Denovium™ Engine learns and becomes a better predictor of drug candidate sequences and cell line designs that will achieve the characteristics we desire.
With a revolutionary machine learning approach and the learnings of a decade collecting data about protein-protein interactions, Absci is developing a computational pipeline for target-directed therapeutic protein design. We believe this technology will have an enormous impact on the biopharma industry as we pursue our ultimate goal – getting better medicines to patients faster.
Using our proprietary experimental datasets and public protein databases, we train our Denovium™ Engine to understand the drug characteristics we care about, and make predictions about how to improve drug candidate sequences. We have shown that our models can accurately predict target binding affinity for antibody sequence variants, as well as score “naturalness,” which is associated with multiple drug developability characteristics. Integrating these models with generative techniques, our Denovium™ Engine allows us to optimize for several features simultaneously to create the best drug candidates.
Our Denovium™ Engine understands not only the determinants for optimizing drug sequences, but also the cell line features of our proprietary SoluPro™ E. coli cell line that will enable robust protein production. With manufacturability being a frequent stumbling block for next-generation biologics, being able to predict cell line designs and efficiently engineer high performance SoluPro™ production strains is an important value driver for our discovery and cell line development programs.
We harness the power of our synthetic biology platform to evaluate millions to billions of protein sequence variants and cell line designs, and the data we amass fuels our Denovium™ Engine. Our models, which we’ve continuously developed and refined over the course of years, enable us to generate, test, and optimize orders of magnitude more potential drug candidates at a fraction of the time and cost of traditional high-throughput screening assays.
Importantly, we validate and refine the AI performance by producing the proteins and engineering the cell lines predicted by our Denovium™ Engine, and testing them in our wet lab, which enables the models to get better and more broadly applicable with each cycle. This positive feedback loop of in silico predictions and wet lab validation data enables us to explore a vast sequence space of next-generation biologics while optimizing for desired drug properties and manufacturing characteristics from the beginning of the discovery process.