SUPR
Machine learning for prediction of protein structure
Dnr:

NAISS 2024/5-554

Type:

NAISS Medium Compute

Principal Investigator:

Andreas Heuer

Affiliation:

Lunds universitet

Start Date:

2024-11-27

End Date:

2025-12-01

Primary Classification:

30105: Neurosciences

Webpage:

Allocation

Abstract

AAV vectors are the current forerunner in delivering transgenes to the cell type of interest. However, the naturally occurring vectors do not have a high tropism (specificity) but infect a range of cells. By rational design (e.g. the capsid structure) it is possible to change the tropism to develop new variants that have different properties. Using machine learning we aim to predict protein receptor interactions. Recent technological advances enable the prediction of protein structures from genetic sequences. In the current project we will train machine learning models to predict the function of new sequences, thereby optimising AAV vector capsids with the goal of finding new variants that possess novel capabilities such as improved specificity or immune evasion.