Enoch’s artificial intelligence identifies high-quality therapeutic formulations for neurodegenerative disease. The cutting-edge machine learning derives dynamic representations of the regulatory networks governing cell behavior. Enoch learns interpretable models from single-cell sequencing and perturbation experiments, enabling them to provide high-confidence predictions for multiple disorders. The machine learning approach is original and custom-built for our problem space. More traditional deep learning methods have been successfully applied to a variety of domains in single cell biology. Enoch goes a step beyond these methods by designing a new neural network architecture that can learn literal representations of the molecular events governing dynamic network behavior.