Background and problem

Artificial intelligence-driven modeling of protein complexes

Proteins often work in pairs or groups known as complexes to accomplish the vital functions of the living cells and organisms. While some of the protein-protein interactions are well studied, most of the therapeutically important protein complexes are structurally unsolved. Constructing an accurate model of these complexes would shed light on many fundamental biological functions. This project aims to build molecular models for therapeutically important human protein complexes. Here, the focus will be on nuclear receptors which are targets of 3% of drugs in the clinic. 

Goal

We will use a combination of deep learning trained models, molecular modeling, and molecular dynamics simulations to accurately construct both homo- and hetero-dimers of nuclear receptors. Later, by validating the models with information from complementary experiments, we will use the models as a new starting point on designing drugs that encourage or discourage receptor dimerization which will be instrumental to develop novel treatments for diseases such as myeloma and inflammation.

Techniques: Molecular modeling, Deep learning-based protein structure prediction methods, Molecular dynamics simulation 

Remarks

III. No use of laboratory animals/geen gebruik van proefdieren

The key prerequisite for the project is the basic knowledge of protein structure and chemistry. The basic knowledge about programming/scripting and about the Linux commands are helpful, but they are not essential.