Membrane analysis toolkit: a quantitative way to analyse protein lipid mixtures from cryo-EM images

Abstract text (incl. figure legends and references)

Cryo-EM is a powerful tool to study protein membrane interactions as it delivers high-resolution images of vitrified macromolecules including biological membranes. The quantitative analysis of protein-induced effects on the membrane structure is a challenging and time-consuming task as they are not amenable to established single-particle image reconstruction techniques. However, due to the high contrast of phospholipid bilayers, the images are suitable for detailed analysis and can yield characteristic lipid features and membrane shapes. Currently, localisation of such membrane structures and the extraction of their features is often performed interactively and is, therefore, subjective and a time-consuming task.

Here we present the "membrane analysis toolkit", a toolkit currently in development which aims to automate the membrane analysis and facilitate quantitative comparison of large functional datasets, e.g capturing the variations such as protein mutations and lipid compositions. The membrane structures are localized using convolutional neural networks and sequentially analysed by a variety of image processing techniques. Typical determined features are membrane thickness, curvature and distance structures along the membranes in addition to shape information about the continuous membrane structures like total area and shape classification.

The membrane analysis toolkit can be used for any sample containing lipid bilayers and can lead to a more quantitative understanding of membrane-active proteins or other environmental influences.