Deep learning for drug discovery
From a massive database of biological experiments, our algorithms learn deep models linking raw chemical structure to physiological function and target activity.
The search for new drugs is still extremely slow and costly. Of 5000 leads that have been identified by researchers, only one has the properties required to become a medication.
Modern machine learning enables us to rapidly and confidently screen and characterize high quality compounds before they enter expensive preclinical trials.
Our system consistently outperforms the latest published research, yet we continuously improve our database, training strategy, and neural network architectures to reach for ever better prediction accuracies.
