Documentation

SuperResNET supports loading SMLM point cloud data from various microscopes (e.g., dSTORM, PALM, DNA-PAINT) with different file formats (e.g., non-text binary, text, and ASCII formats). See the microscopy data formats that SuperResNET supports in the video

Note 1. SuperResNET will inform you when trying to load unsupported file format. You might need to prepare your file into one of the supported microscopy formats (e.g., .3dlp, .bin, .ascii) or generate a file with the localization coordinates in X, Y, Z and save it as one of the supported general formats (e.g., .txt, .xyz). You can save the file with or without header information. SuperResNET can load both types of files

Please contact  Ismail Khater if you have trouble loading your data

See the step-by-step video guide for using SuperResNET or read the details below

Use this tab to alleviate the multiple blinking effects of a single fluorophore artifact by selecting a proper merging threshold [1]. Select the merging threshold = 0 nm then click on the Merge button to skip this multiple blinking correction step and only remove the duplicate localizations that share the exact same coordinates. Next, select a proximity threshold (PT) to construct a network graph from the localizations and calculate the degree of the localizations. Select the PT parameter according to the size of clustering then click on the Calculate button. SuperResNET visualizes the histograms of the network degree before and after merging as well as the localizations before and after the merging.

Optionally, calculate Ripley’s H-function to provide a heuristic for setting PT. Ripley's H-function can be used to return the global scale for the clustering (r).  As a rule of thumb, we can construct a network to extract the homogenous clusters in the data by using a proximity threshold PT as (r/2 ≤ PT ≤  r ). For heterogeneous clusters, PT  ≅ r/2.

Calculating Ripley's H-function may take a long time to compute depending on the r_min and r_max parameters and the size of the data. 

Ripley's H-function can also provide insight for setting the kernel bandwidth when segmenting the blobs using mean-shift (see the Segment Tab).

Note 1:  In SuperResNET Ripley’s H-function is calculated for the 2D projected SMLM data because the 3D data has a lateral resolution of ~20 nm and an axial resolution of ~40-50 nm. Don’t use this Ripley’s H-function if you have deep Z (e.g. more than 1 µm).

Note 2: Every time you use a new merging threshold value and click on the Merge button, you have to click on the Draw button after it to find the new parameters.

[1] Khater IM, Meng F, Wong TH, Nabi IR, Hamarneh G. Super resolution network analysis defines the molecular architecture of caveolae and caveolin-1 scaffolds. Scientific reports. 2018 Jun 13;8(1):9009.

The next module is noise filtering. Click on the Filter button to calculate the filtering parameter [1]. After that, the user can move the Alpha Slider to control noise removal in real-time by noticing both the scatter plot and the histograms.

[1] Khater IM, Meng F, Wong TH, Nabi IR, Hamarneh G. Super resolution network analysis defines the molecular architecture of caveolae and caveolin-1 scaffolds. Scientific reports. 2018 Jun 13;8(1):9009.

The aim of the segmentation is to find the membership of every localization to one and only one cluster or blob. The Segment module supports two methods for segmenting the SMLM localizations:

1. The mean-shift algorithm [1] [2] is suitable for segmenting blob-like structures. It requires the user to specify the kernel bandwidth. As a rule of thumb, select the kernel bandwidth to be less than or equal to the r value found using Ripley’s H-function (see the Merge tab above).  

2. DBSCAN [3] is suitable for segmenting clusters with non-spherical-like shapes. However, it requires setting two parameters, Epsilon and Min. Pnts. Setting both parameters depends on the size of the clusters, spacing between the clusters, density of the clusters, and the amount of noise. It may not always be straightforward to set both parameters for the desired results. 

After selecting the segmentation method, click on the Segment button. Optionally, set the Min. Blob Size to visually filter the clusters/blobs according to their number of localizations.

[1] Fukunaga K, Hostetler L. The estimation of the gradient of a density function, with applications in pattern recognition. IEEE Transactions on information theory. 1975 Jan;21(1):32-40.

[2] Comaniciu D, Meer P. Mean shift: A robust approach toward feature space analysis. IEEE Transactions on Pattern Analysis & Machine Intelligence. 2002 May 1(5):603-19.

[3] Ester M, Kriegel HP, Sander J, Xu X. A density-based algorithm for discovering clusters in large spatial databases with noise. InKdd 1996 Aug 2 (Vol. 96, No. 34, pp. 226-231).

The Blob Features module used to extract a set of features/descriptors for every segmented blob. SuperResNET extracts 28 features that capture the shape, size, volume, network measures, etc. for every blob [1]. The user can show the histogram for some of the extracted features by selecting the feature name from the drop-down component.

[1] Khater IM, Meng F, Wong TH, Nabi IR, Hamarneh G. Super resolution network analysis defines the molecular architecture of caveolae and caveolin-1 scaffolds. Scientific reports. 2018 Jun 13;8(1):9009.

Unsupervised machine learning can estimate the groups in the data based on their features. SuperResNET uses the K-means method for grouping the blobs. Set the number K (an integer between  1 and 14) of groups (or classes) of blobs then click Group.

SuperResNET implements the Silhouette Criterion, which can help the user evaluate the number of groups when clustering the blobs. Calculating the Silhouette Criterion uses Linkage as the clustering algorithm and the correlation (one minus the sample correlation between points) as the distance metric.

This module provides interactive visualization of blobs.  Sort/filter the blobs based on a selected feature and visualize the blob localizations and the blob network by clicking on the row of the table. Scroll the blobs in the table and select them for visualizations. 

For example, retrieve the most spherical or linear blobs and visualize them. We show an example of visualizing the most spherical blob. Move the sliders and edit fields to change the transparency, the localization/node size, proximity threshold/connectivity of the localizations, and the shrinkage factor for the boundary visualization.

This module used to do more analysis for the selected blob. The modularity analysis allows studying the modules of interacting molecules within the blob [1]. The user can interactively visualize the modules in various ways.

[1] Khater IM, Liu Q, Chou KC, Hamarneh G, Nabi IR. Super-resolution modularity analysis shows polyhedral caveolin-1 oligomers combine to form scaffolds and caveolae. Scientific reports. 2019 Jul 8;9(1):1-0.

This module is used for finding representative blobs from every group. Specify the number of blobs with features most similar to the mean features of every individual group. These top blobs are then overlaid and visualized in various ways.  

In the About tab, you can update your license key, as described in the Installation documentation.  The About tab also includes contact information. We are pleased to receive your questions and feedback about SuperResNET. Please contact Ismail Khater