The ECoSystem
Precision medicine through polymathic AI
Representing biology at higher levels of abstraction requires a full tech stack approach
We are vertically integrating many component parts of drug discovery into a complex coordination model. We believe we cannot outsource anything mission-critical without sacrificing quality, so we have unified everything we need in house to discover drugs for patients with brain diseases.
THE CORE
Multimodal data generation combined with biology and chemistry foundation models
%20opt1%20(1).png)
Digitising Biology
Hyper-scale single-cell dataset derived from patient tissue
We have created a single-cell CRISPRi perturbation dataset that is larger than all other published data sets in the world combined (1). We grow cells from patient surgical resections in 3D, allowing us to recreate the clinical situation as closely as possible. To date, we have perturbed the entire functional genome in millions of single cells in thousands of genetic contexts. The tools we built to generate this dataset are scalable across many brain diseases.
%20opt1%20(1).png)
Biology foundation models
Exploiting the vulnerability of disease to identify precision drug targets
Our dataset was custom-built to train our biology foundation models, which allows us to predict the association between gene perturbation and transcriptomic-phenotypic responses within a given genetic context. With this capability, we can identify precision drug targets and companion genetic biomarkers.
%20opt1%20(2).png)
Chemistry foundation models
From drug targets to drug compounds
We enumerate in silico chemical libraries of over a trillion compounds to find new drugs for our targets. Our chemistry foundation models have been pre-trained on over 37 billion chemical structures, which allows us to accelerate our path to clinic.
Sources