Quality-driven single particle averaging of super-resolved sarcomeres

Session

Reproducibility of Data Analysis at Scale

Authors

Alicia Cuber Caballero (1), Dr. Pauline Bennett (1), Dr. Richard Marsh (1), Prof. Mathias Gautel (1), Dr. Sian Culley (1)

Affiliations

1. King's College London

Keywords

super-resolution, image analysis, STORM, STED, single particle averaging, muscle biology

Abstract text

Sarcomeres are the base contractile unit in muscle tissue and are symmetrical, crystalline structures; in the nanoscale, their structure and organisation has not been fully resolved. The M-band is the central region of sarcomeres, and  is characterised by high molecular density within a small spatial region (~100nm). Many of the proteins within this region form doublets, and visualising the separation of these requires high resolution microscopy. Electron microscopy methods are limited due to sharing of domains and motifs between neighbouring proteins, and as a result super-resolution techniques such as STED and STORM offer a promising alternative. However, conventional antibody labelling can have difficulty penetrating these dense structures and can introduce linker errors due to the dual-antibody complex required for indirect immunostaining.  

For these reasons, there is a lot of variability in super-resolution images of sarcomere doublet structures even within a single field of view. For example, some will not be perfectly parallel due to mechanical stresses during sample preparation or localization biases in STORM. Furthermore, label coverage is often incomplete, making these lines appear ‘broken’ or ‘dotty’. These issues exemplify the need for robust computational approaches that can provide confidence measuring the separation of lines within doublets and improve their accuracy, especially when optimising super-resolution sample preparation and acquisition is time-consuming and challenging. 

To investigate factors affecting the accuracy of doublet separation measurements, we use both simulated and real data. We have developed a pipeline, which identifies and extracts doublets from an image, assesses each doublet´s quality, and then implements a single particle averaging approach weighted by these qualities. Initially, we validate this approach using simulated data and super-resolution images of doublets with well-known separations. Going forward, this pipeline will be a valuable tool for benchmarking novel sarcomere labelling approaches (such as nanobodies), super-resolution modalities (such as expansion microscopy) and retrieving confident measurements of sarcomere protein separations that have not previously been characterized. 

References