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

Important results from the project

Our objective was to create a solution so that researchers at Swedish companies can access Sweden´s largest supercomputer through their browser with just one click. Within two months, we had set up the first prototype. Through it, partner AstraZeneca got access to a newly developed software from KTH. This was adapted to create data for molecules, and with that data a predictive AI model was created. Several other companies are very interested in participating in the next phase of the project.

Expected long term effects

In the next phase, we want to use the supercomputer to create data, store experimental data and train more advanced AI models on the data. These AI-models will be used as so-called agents, computers that make decisions. They will therefore be made by models on the supercomputer about how new experiments should be designed, experiments that will be carried out entirely by robots. This will lead to an extremely much higher rate of new discoveries, which can lead to, for example, new drugs faster.

Approach and implementation

We assembled a team of system administrators, researchers in theoretical chemistry, medicine and simulations, and experimental chemists. The project began by setting a goal for what the technical solution would look like. When the first version was in place, researchers at KTH created a customized software that was then used by researchers at AstraZeneca. In parallel, we have developed the idea for the next phase and been in contact with other actors.