Interactive segmentation of membrane and membrane mimic densities in cryo-EM maps

Abstract

Contextual density arising from biological membranes or membrane mimics such as detergent micelles or lipid nanodiscs, is ubiquitous in cryo-EM maps of integral and peripheral membrane proteins, or membrane-associated complexes, from single-particle analysis and subtomogram averaging. Here, we describe SURFER, a lightweight, GPU-accelerated extension for UCSF ChimeraX that enables rapid semantic differentiation of membrane or membrane mimic density from macromolecular signal. SURFER operates on unfiltered maps to automatically generate segmentation masks that can be flexibly applied to any aligned target map, including reference volumes for local refinement or post-processed maps optimised for visualisation and interpretation. By allowing contextual density to be selectively toggled or subtracted, SURFER enables direct interactive comparison between representations with and without membrane-related features to facilitate consistent analysis and reporting of membrane context in cryo-EM structures. An interactive graphical interface within ChimeraX supports convenient threshold adjustment, connectivity filtering, and immediate visual evaluation of the resulting segmentation.