A framework for evaluating the performance of SMLM cluster analysis algorithms

Distinguishment of different varieties of rhubarb based on UPLC fingerprints and chemometrics

Rhubarb, a widely used traditional Chinese medicine (TCM), is primarily used for purging in practice. It is derived from the dried roots and rhizomes of R. tanguticum Maxim. ex Balf. (RT), Rheum officinale Baill. (RO) and R. palmatum L. (RP). To date, although the three varieties of rhubarb have been used as the same medicine in clinical, studies have found that they have different chemical compositions and pharmacological effects. To ensure the stability of rhubarb for clinical use, a simple and effective method should be built to compare and discriminate three varieties of rhubarb. Here, ultra-performance liquid chromatography-diode array detection (UPLC-DAD) fingerprints combined with chemometric methods were developed to evaluate and discriminate 29 batches of rhubarb. Similarity evaluation, hierarchical cluster analysis (HCA) and principal component analysis (PCA) showed that the chemical constituents of the three varieties of rhubarb were significantly different, and the three varieties could be effectively distinguished. Finally, all the 14 common peaks were identified by ultra-performance liquid chromatography coupled with quadrupole-time-of-flight mass spectrometry (UPLC-Q-TOF-MS). In this research, the developed UPLC fingerprints offer a simple, reliable and specific approach for distinguishing different varieties of rhubarb. This research aims to promote the scientific and appropriate clinical application of rhubarb from three varieties.

SEMORE: SEgmentation and MORphological fingErprinting by machine learning automates super-resolution data analysis

The morphology of protein assemblies impacts their behaviour and contributes to beneficial and aberrant cellular responses. While single-molecule localization microscopy provides the required spatial resolution to investigate these assemblies, the lack of universal robust analytical tools to extract and quantify underlying structures limits this powerful technique. Here we present SEMORE, a semi-automatic machine learning framework for universal, system- and input-dependent, analysis of super-resolution data. SEMORE implements a multi-layered density-based clustering module to dissect biological assemblies and a morphology fingerprinting module for quantification by multiple geometric and kinetics-based descriptors. We demonstrate SEMORE on simulations and diverse raw super-resolution data: time-resolved insulin aggregates, and published data of dSTORM imaging of nuclear pore complexes, fibroblast growth receptor 1, sptPALM of Syntaxin 1a and dynamic live-cell PALM of ryanodine receptors. SEMORE extracts and quantifies all protein assemblies, their temporal morphology evolution and provides quantitative insights, e.g. classification of heterogeneous insulin aggregation pathways and NPC geometry in minutes. SEMORE is a general analysis platform for super-resolution data, and being a time-aware framework can also support the rise of 4D super-resolution data.

Single-molecule localization microscopy reveals STING clustering at the trans-Golgi network through palmitoylation-dependent accumulation of cholesterol

Stimulator of interferon genes (STING) is critical for the type I interferon response to pathogen- or self-derived DNA in the cytosol. STING may function as a scaffold to activate TANK-binding kinase 1 (TBK1), but direct cellular evidence remains lacking. Here we show, using single-molecule imaging of STING with enhanced time resolutions down to 5 ms, that STING becomes clustered at the trans-Golgi network (about 20 STING molecules per cluster). The clustering requires STING palmitoylation and the Golgi lipid order defined by cholesterol. Single-molecule imaging of TBK1 reveals that STING clustering enhances the association with TBK1. We thus provide quantitative proof-of-principle for the signaling STING scaffold, reveal the mechanistic role of STING palmitoylation in the STING activation, and resolve the long-standing question of the requirement of STING translocation for triggering the innate immune signaling.

Biological Applications of Synthetic Binders Isolated from a Conceptually New Adhiron Library

BACKGROUND

Adhirons are small (10 kDa) synthetic ligands that might represent an alternative to antibody fragments and to alternative scaffolds such as DARPins or affibodies.

METHODS

We prepared a conceptionally new adhiron phage display library that allows the presence of cysteines in the hypervariable loops and successfully panned it against antigens possessing different characteristics.

RESULTS

We recovered binders specific for membrane epitopes of plant cells by panning the library directly against pea protoplasts and against soluble C-Reactive Protein and SpyCatcher, a small protein domain for which we failed to isolate binders using pre-immune nanobody libraries. The best binders had a binding constant in the low nM range, were produced easily in bacteria (average yields of 15 mg/L of culture) in combination with different tags, were stable, and had minimal aggregation propensity, independent of the presence or absence of cysteine residues in their loops.

DISCUSSION

The isolated adhirons were significantly stronger than those isolated previously from other libraries and as good as nanobodies recovered from a naïve library of comparable theoretical diversity. Moreover, they proved to be suitable reagents for ELISA, flow cytometry, the western blot, and also as capture elements in electrochemical biosensors.

Nanoscale CAR Organization at the Immune Synapse Correlates with CAR-T Effector Functions

T cells expressing chimeric antigen receptors (CARs) are at the forefront of clinical treatment of cancers. Still, the nanoscale organization of CARs at the interface of CAR-Ts with target cells, which is essential for TCR-mediated T cell activation, remains poorly understood. Here, we studied the nanoscale organization of CARs targeting CD138 proteoglycans in such fixed and live interfaces, generated optimally for single-molecule localization microscopy. CARs showed significant self-association in nanoclusters that was enhanced in interfaces with on-target cells (SKOV-3, CAG, FaDu) relative to negative cells (OVCAR-3). CARs also segregated more efficiently from the abundant membrane phosphatase CD45 in CAR-T cells forming such interfaces. CAR clustering and segregation from CD45 correlated with the effector functions of Ca++ influx and target cell killing. Our results shed new light on the nanoscale organization of CARs on the surfaces of CAR-Ts engaging on- and off-target cells, and its potential significance for CAR-Ts' efficacy and safety.

clusterBMA: Bayesian model averaging for clustering

Various methods have been developed to combine inference across multiple sets of results for unsupervised clustering, within the ensemble clustering literature. The approach of reporting results from one 'best' model out of several candidate clustering models generally ignores the uncertainty that arises from model selection, and results in inferences that are sensitive to the particular model and parameters chosen. Bayesian model averaging (BMA) is a popular approach for combining results across multiple models that offers some attractive benefits in this setting, including probabilistic interpretation of the combined cluster structure and quantification of model-based uncertainty. In this work we introduce clusterBMA, a method that enables weighted model averaging across results from multiple unsupervised clustering algorithms. We use clustering internal validation criteria to develop an approximation of the posterior model probability, used for weighting the results from each model. From a combined posterior similarity matrix representing a weighted average of the clustering solutions across models, we apply symmetric simplex matrix factorisation to calculate final probabilistic cluster allocations. In addition to outperforming other ensemble clustering methods on simulated data, clusterBMA offers unique features including probabilistic allocation to averaged clusters, combining allocation probabilities from 'hard' and 'soft' clustering algorithms, and measuring model-based uncertainty in averaged cluster allocation. This method is implemented in an accompanying R package of the same name. We use simulated datasets to explore the ability of the proposed technique to identify robust integrated clusters with varying levels of separation between subgroups, and with varying numbers of clusters between models. Benchmarking accuracy against four other ensemble methods previously demonstrated to be highly effective in the literature, clusterBMA matches or exceeds the performance of competing approaches under various conditions of dimensionality and cluster separation. clusterBMA substantially outperformed other ensemble methods for high dimensional simulated data with low cluster separation, with 1.16 to 7.12 times better performance as measured by the Adjusted Rand Index. We also explore the performance of this approach through a case study that aims to identify probabilistic clusters of individuals based on electroencephalography (EEG) data. In applied settings for clustering individuals based on health data, the features of probabilistic allocation and measurement of model-based uncertainty in averaged clusters are useful for clinical relevance and statistical communication.

Optimizing Voronoi-based quantifications for reaching interactive analysis of 3D localizations in the million range

Over the last decade, single-molecule localization microscopy (SMLM) has revolutionized cell biology, making it possible to monitor molecular organization and dynamics with spatial resolution of a few nanometers. Despite being a relatively recent field, SMLM has witnessed the development of dozens of analysis methods for problems as diverse as segmentation, clustering, tracking or colocalization. Among those, Voronoi-based methods have achieved a prominent position for 2D analysis as robust and efficient implementations were available for generating 2D Voronoi diagrams. Unfortunately, this was not the case for 3D Voronoi diagrams, and existing methods were therefore extremely time-consuming. In this work, we present a new hybrid CPU-GPU algorithm for the rapid generation of 3D Voronoi diagrams. Voro3D allows creating Voronoi diagrams of datasets composed of millions of localizations in minutes, making any Voronoi-based analysis method such as SR-Tesseler accessible to life scientists wanting to quantify 3D datasets. In addition, we also improve ClusterVisu, a Voronoi-based clustering method using Monte-Carlo simulations, by demonstrating that those costly simulations can be correctly approximated by a customized gamma probability distribution function.

Introducing Diinamic, a flexible and robust method for clustering analysis in single-molecule localization microscopy

Super-resolution microscopy allowed major improvements in our capacity to describe and explain biological organization at the nanoscale. Single-molecule localization microscopy (SMLM) uses the positions of molecules to create super-resolved images, but it can also provide new insights into the organization of molecules through appropriate pointillistic analyses that fully exploit the sparse nature of SMLM data. However, the main drawback of SMLM is the lack of analytical tools easily applicable to the diverse types of data that can arise from biological samples. Typically, a cloud of detections may be a cluster of molecules or not depending on the local density of detections, but also on the size of molecules themselves, the labeling technique, the photo-physics of the fluorophore, and the imaging conditions. We aimed to set an easy-to-use clustering analysis protocol adaptable to different types of data. Here, we introduce Diinamic, which combines different density-based analyses and optional thresholding to facilitate the detection of clusters. On simulated or real SMLM data, Diinamic correctly identified clusters of different sizes and densities, being performant even in noisy datasets with multiple detections per fluorophore. It also detected subdomains ("nanodomains") in clusters with non-homogeneous distribution of detections.

Super-resolved trajectory-derived nanoclustering analysis using spatiotemporal indexing

Single-molecule localization microscopy techniques are emerging as vital tools to unravel the nanoscale world of living cells by understanding the spatiotemporal organization of protein clusters at the nanometer scale. Current analyses define spatial nanoclusters based on detections but neglect important temporal information such as cluster lifetime and recurrence in "hotspots" on the plasma membrane. Spatial indexing is widely used in video games to detect interactions between moving geometric objects. Here, we use the R-tree spatial indexing algorithm to determine the overlap of the bounding boxes of individual molecular trajectories to establish membership in nanoclusters. Extending the spatial indexing into the time dimension allows the resolution of spatial nanoclusters into multiple spatiotemporal clusters. Using spatiotemporal indexing, we found that syntaxin1a and Munc18-1 molecules transiently cluster in hotspots, offering insights into the dynamics of neuroexocytosis. Nanoscale spatiotemporal indexing clustering (NASTIC) has been implemented as a free and open-source Python graphic user interface.

Error-Correction Method for High-Throughput Sizing of Nanoscale Vesicles with Single-Molecule Localization Microscopy

Single-molecule localization microscopy (SMLM) allows super-resolution imaging, mapping, counting, and sizing of biological nanostructures such as cell organelles and extracellular vesicles (EVs), but sizing structures smaller than ∼100 nm can be inaccurate due to single-molecule localization error caused by distortion of the point spread function and limited photon number. Here we demonstrate a method to correct localization error when sizing vesicles and other spherical nanoparticles with SMLM and compare sizing results using two vesicle labeling schemes. We use mean approximation theory to derive a simple equation using full width at half-maximum (FWHM) for correcting particle sizes measured by two-dimensional SMLM, validate the method by sizing streptavidin-coated polystyrene nanobeads with the SMLM technique dSTORM with and without error correction, using transmission electron microscopy (TEM) for comparison, and then apply the method to sizing small seminal EVs. Nanobead sizes measured by dSTORM became increasingly less accurate (larger than TEM values) for beads smaller than 50 nm. The error-correction method reduced the size difference versus TEM from 15% without error correction to 7% with error correction for 40 nm beads, from 44% to 9% for 30 nm beads, and from 66% to 15% for 20 nm beads. Seminal EVs were labeled with a lipophilic membrane dye (MemBright 700) and with an Alexa Fluor 488-anti-CD63 antibody conjugate, and were sized separately using both dyes by dSTORM. Error-corrected exosome diameters were smaller than uncorrected values: 72 nm vs 79 nm mean diameter with membrane dyes; 84 nm vs 97 nm with the antibody-conjugated dyes. The mean error-corrected diameter was 12 nm smaller when using the membrane dye than when using the antibody-conjugated dye likely due to the large size of the antibody. Thus, both the error-correction method and the compact membrane labeling scheme reduce overestimation of vesicle size by SMLM. This error-correction method has a low computational cost as it does not require correction of individual blinking events, and it is compatible with all SMLM techniques (e.g., PALM, STORM, and DNA-PAINT).