Description
Biological membranes, especially organelle membranes, are highly complex structures composed of diverse proteins and lipid types that are often distributed unevenly across the membrane surface. Understanding how these molecules are arranged laterally is essential for revealing how organelles interact with their environment. Although extensive experimental data exist on global membrane composition, obtaining detailed information about the lateral organization of membrane components remains challenging. Molecular dynamics (MD) simulations offer one possible route to study this organization. However, due to their high computational cost, conventional MD techniques struggle to achieve the precision and scale needed to model entire organelle membranes. Therefore, this project aims to produce a "computational microscope" that integrates machine learning (ML) to predict the lipids organization in the membrane. By training ML models on data from simulations of smaller membrane patches consisting of the structural properties of the membrane, we aim to establish rules that allow these local predictions to be combined into a coherent model of the full membrane. Such an approach could significantly enhance the efficiency of MD simulations of cell scale system by generating improved initial configurations for membrane molecules, ultimately accelerating studies of membrane structure and function.
| Field of study | Biophysics |
|---|---|
| Supervisor | Weria Pezeshkian and Julius Bier Kierkegaard |