Cloud based integration of advanced software for data generation and prediction in drug development

Purpose and goal

By creating a cloud-based solution connected to supercomputers, we make the resources available to more users. A supercomputer will be a click away via a web browser. The resources will feature newly developed world-leading software and tailored workflows. These will be used for large-scale data generation for training AI models in drug development and for prediction of molecular properties.

Expected effects and result

A prototype for the cloud service will be created. We will evaluate the user interface so that everything is perceived as seamless and smooth for the user. Several workflows will be tested where the newly developed software VeloxChem is used to generate data coupled to molecular properties that are of great importance for successful drug development. In this way, Sweden´s most powerful supercomputers are made available to more users and we take modern software from academia to industry.

Planned approach and implementation

At KTH, the cloud solution will be created and first tested internally. Once a working version is ready, AstraZeneca users will be given access. Specific modules needed for AstraZeneca´s drug development will be created in VeloxChem and workflows based on Python will be created at KTH. AstraZeneca is responsible for generating molecular input data for the models and for evaluating the setup. The data will be tested in graph-based neural networks.