Technology

By leveraging recent advances in transfer-learning, our platform requires significantly less training data than existing approaches, in some cases as few as 10 training samples, removing access to data as a principle rate limiting factor when deploying machine-learned solutions in pharmaceutical R&D.


Why does this matter?

It now takes 10-15 years and almost $2B to bring a new drug to market, largely due to the fact that ~90% of compounds fail in clinical testing.

We believe that novel machine learning algorithms, properly designed and appropriately applied, have the potential to provide a solution.

Technology Overview

Novel approaches to generative drug design, virtual screening, and AI-driven lead optimization

Modules in Development

Clinical development

Clinical Development Module

Predict a compound’s likelihood of success in a specific patient population, allowing for rapid filtering and prioritization of leads with the highest probabilities of eventual success.

Preclinical toxicology

Virtual Screening Module

Simulate a broad selection of cell-based functional assays to better understand relevant ADMET properties early in the discovery process.

Drug-Drug interactions

Drug-Drug Interactions Module

Inform preclinical safety assessments by providing an exhaustive overview of potential adverse drug-drug interactions.

Metabolite

Generative Module

Generate novel drug candidates that are simultaneously optimized for multiple molecular properties (including toxicity, ADME, efficacy, bioavailability, and synthesizability, among others).

Explainer Functionality

Provide informed decision making through explainable and interpretable models that provide insight into which molecular fragments or pathway components are driving the predicted outcomes.

Enhancer Functionality

Propose molecular modifications to increase binding affinity while minimizing off-target interactions and optimizing pharmacological characteristics.

Line
E 2

Drive informed lead optimization by using the explainer module to identify negative critical fragments before subsequent optimization.