Single Molecule Localisation Microscopy (SMLM) is increasingly viewed as one of the major tool for analysis of biological processes on a high resolution level in the range of 10 to 50 nm. The procedure relies on sequential detection of (a subset of) individual fluorophores. For dense regions (fluorophores with significant overlap), a compromise between fluorescence labelling density and the photoswitching behaviour of fluorophores is needed to have an optical isolation i.e. sparse distribution of molecules in each acquired frame.
Algorithms used to precisely identify the locations of these fluorophores can be broadly classified into two categories, namely fitting based and non-fitting based (usually Centroid) methods. While iterative fitting-based methods can usually provide fitted parameters equal or close to the maximum likelihood estimate, ad hoc centroid based methods are usually very quick. However, all localisation methods struggle if the underlying model poorly represents the observed data e.g. background level, out of focus signals, noise, etc. A particular challenge for the exact fluorophore determination is posed by spatially as well as temporally fluctuating background intensities arising from out of focus blinking fluorophores. This is to some degree always given if the structure is not per se 2-dimensional (e.g. PALM using TIRF illumination).
Here, we present a comparative analysis of a range of available localisation algorithms on complexity, applicability and performance by testing them on both synthetic and experimental data that cover examples of both sparse and dense regions, with both low and high background levels to determine, which method is suited for a given set of data.
We gratefully acknowledge the colleagues at IMB who supported us with reagents. In particular, we would like to thank Aleksander Szczurek and Hyun-Keun Lee for samples, reagents and many interesting discussions. This work is supported by the Boehringer Ingelheim Foundation. The support of University Hospital Heidelberg (Prof. S. Dithmar) to G.B. and M.H. is also gratefully acknowledged.