TY  - GEN
N2  - Super-resolution techniques have enabled fluorescence microscopy to surpass the diffraction limit of light and resolve nanoscale biological structures with molecular specificity. As with other microscopy data, quantitative analyses of super-resolution images have enabled great insight into the underlying architecture of many macromolecular structures. However, this is still a challenging process, especially in the field of single-molecule localization microscopy (SMLM). Unlike pixelated images yielded by most microscopy techniques, SMLM data is composed of a list of fluorophore coordinates and their specific positional uncertainties. Therefore, applying pixel-based approaches to SMLM data requires image rendering, which can cause loss of information and can complicate the analysis. These drawbacks can be mitigated by using coordinate-based approaches.
However, currently available coordinate-based approaches in SMLM are typically only applicable to simple geometries or require identical structures. These approaches do not support a basic task in structural analysis: postulating a model with a suitable underlying geometry to probe key parameters in a biological structure.
Here, I present LocMoFit (Localization Model Fit), a new framework for fitting a parameterized geometric model to SMLM data at the level of localizations. Based on maximum likelihood estimation, LocMoFit extracts meaningful parameters from individual structures and can select the most suitable model.
Using the nuclear pore complex, microtubules, and clathrin-mediated endocytosis (CME) as examples, I demonstrate the application of LocMoFit in in situ structural biology for extracting descriptive parameters of complex, heterogeneous and even dynamic structures. Beyond that, I further showcase applications including assembling multi-protein distribution maps of six nuclear pore components, calculating single-particle averages without any structural prior, and reconstructing the progression of endocytosis - a highly dynamic process - from static snapshots.
On the one hand, the quantitative analysis allowed to address a long-standing controversy of how the clathrin coat is rearranged during CME in mammalian cells. On the other hand, the dynamic reconstruction of representative endocytic proteins over the endocytic progression shows the potential of revealing previously unknown nanoscale features that may be associated with force generation during CME in yeast.
To ensure all these functionalities are accessible, I implemented LocMoFit as open-source and provided instructions and model templates. A simulation engine and visualization routines are also supplied for users to examine the plausibility of their own analysis, which is also what I used to validate the robustness of the framework in this work. With these, I believe LocMoFit will enable any user to extend the information that can be faithfully extracted from SMLM data.
A1  - Wu, Yu-Le
AV  - public
Y1  - 2022///
UR  - https://archiv.ub.uni-heidelberg.de/volltextserver/31920/
ID  - heidok31920
CY  - Heidelberg
TI  - Maximum-likelihood model fitting for quantitative analysis of SMLM data
ER  -