A Semi-supervised Learning Approach for Calculation of Membrane Curvature Properties, with Application to Mitochondrial Model Membranes

Abstract

Curvature is an important feature of biological membranes that plays a diverse role in regulating vital cellular processes including oxidative phosphorylation, vesicle trafficking, and membrane remodeling. Lipid composition can modulate membrane curvature, particularly in the inner mitochondrial membrane (IMM), where locally enriched nanodomains of cardiolipin have been shown to spontaneously generate local negative mean curvature. Molecular dynamics simulations of highly curved membranes can give key insights into the interplay between membrane curvature and biology, but there is currently a lack of reliable high-performance methods for calculating curvature properties of large membranes with many thousands of lipids.