Artificial intelligence for protein stabilisation in vaccine design

The Informatics Institute of the University of Amsterdam and Janssen R&D Vaccines team up to create an artificial intelligence that can improve current vaccine.

The necessity for quick reliable vaccine development has recently been brought to widespread attention due to the COVID-19 pandemic. New and fast evolving virus variants continuously ignite new pandemic waves. This led to many deaths and considerable strain on the health care systems around the globe. Mutations and the ability of the virus to eventually avoid the mechanisms of the current vaccines call for a quick reliable vaccine responses. However, classical exhaustive experimental approaches are costly and slow. The team will exploit recent advances in artificial intelligence to create an algorithm that may help in identifying effective changes in protein structures for improved vaccine designs. This could result in shortening of the long and costly discovery process.

The efficacy of viral vaccines based on spike (fusion) proteins may depend on the conformation of the spike that is presented to the immune system. Those important proteins tend to contain instable structural parts that might result in undesired structure of the vaccine immunogen. The approach of this project is to create an artificial intelligence that learns from big data sets of known stable and unstable proteins to help identifying unstable structural parts. After such training the AI may then be applied to new viruses which may assist in developing fast, effective, and novel vaccine designs, allowing for quicker vaccine response. 

In this project we will advance recent developments in artificial intelligence, especially deep neural networks, to create and publish an algorithm for the highly accurate detection of unstable protein regions, for use in vaccine design. This will be tested and applied to public and internal protein data sets.

The project will be listed on a subpage here.