StormGraph: A graph-based algorithm for quantitative clustering analysis of diverse single-molecule localization microscopy data

  1. Joshua M. Scurll
  2. Libin Abraham
  3. Da Wei Zheng
  4. Reza Tafteh
  5. Keng C. Chou
  6. Michael R. Gold
  7. Daniel Coombs

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Clustering of proteins is crucial for many cellular processes and can be imaged at nanoscale resolution using single-molecule localization microscopy (SMLM). Ideally, molecular clustering in regions of interest (ROIs) from SMLM images would be assessed using computational methods that are robust to sample and experimental heterogeneity, account for uncertainties in localization data, can analyze both 2D and 3D data, and have practical computational requirements in terms of time and hardware. While analyzing surface protein clustering on B lymphocytes using SMLM, we encountered limitations with existing cluster analysis methods. This inspired us to develop StormGraph, an algorithm using graph theory and community detection to identify clusters in heterogeneous sets of 2D and 3D SMLM data while accounting for localization uncertainties. StormGraph generates both multi-level and single-level clusterings and can quantify cluster overlap for two-color SMLM data. Importantly, StormGraph automatically determines scale-dependent thresholds from the data using scale-independent input parameters. This makes identical choices of input parameter values suitable for disparate ROIs, eliminating the need to tune parameters for different ROIs in heterogeneous SMLM datasets. We show that StormGraph outperforms existing algorithms at analyzing heterogeneous sets of simulated SMLM ROIs where ground-truth clusters are known. Applying StormGraph to real SMLM data in 2D, we reveal that B-cell antigen receptors (BCRs) reside in a heterogeneous combination of small and large clusters following stimulation, which suggests for the first time that two conflicting models of BCR activation are not mutually exclusive. We also demonstrate application of StormGraph to real two-color and 3D SMLM data.

Article activity feed

  1. eLife

    Reviewer #2:

    In their paper "A graph-based algorithm called StormGraph for cluster analysis of diverse single-molecule localization microscopy data", Scurll et al. present a new algorithm to identify clusters in single-molecule localization microscopy (SMLM) data. They use graph-based clustering and show that StormGraph outperforms a selection of existing algorithms, both on simulated and experimental data. The improvement seems not huge, but is convincing, thus this work presents an important contribution to the field. Naturally, not all competing algorithms could be benchmarked in comparison to StormGraph, thus it is not clear if this algorithm is indeed among the best performing algorithms. This is especially true for the cross-correlation analysis. If the applicability of the software included with the manuscript was extended to more potential users, this could be a useful contribution to the field. The manuscript is well written, but quite long. The information content would not be jeopardized if part of the main text and some figures were to be moved to the supplementary information or methods section.

  2. eLife

    Reviewer #1:

    Single molecule localization microscopy (SMLM) has become an important method for understanding the subcellular distribution of fluorescently labelled biomolecules at length scales of a few tens of nanometers. A critical challenge has been to find out, whether and to what extent biomolecular clustering occurs. While methods have been published which address the problem of identifying biomolecular clusters in SMLM images, they still suffer from many user-defined parameters, which - if selected inappropriately - influence the obtained results substantially. The StormGraph-3D method proposed here addresses these issues, based on a comprehensive mathematical framework which reduces the number of user-defined input parameters. The method was evaluated using comprehensive simulations of data, which show its robustness compared to alternative approaches.

    The methods part of the paper would benefit, however, from more realistic data of single molecule blinking behavior, and the evaluation of the consequences on the performance of the method. As the authors acknowledge, overcounting due to blinking has challenged data analysis previously, and gave rise to artifactual localization clusters that do not represent the underlying protein distribution. It would be of particular interest, which results in the method yielded for a random biomolecular distribution.

  3. eLife

    Reviewer #3:

    The authors present the algorithm clearly by comparing it to the most popular SMLM clustering algorithms and showing its robustness in varying density SMLM data, which is a big problem in the field. The presented experimental test on 3D LAMP-1 SMLM data also contributes to the robustness of the paper.

    While reading the manuscript, I missed a comparison with another graph-based SMLM clustering algorithm published previously by Khater et al. in relation to accuracy and computation speed, which is particularly important to demonstrate the advantages of StormGraph. The approach should also be included in Table 1. I also think that a direct comparison in terms of accuracy and computation speed is crucial.

    During the review process, a similar paper has been posted to bioRxiv dated 22. December, https://www.biorxiv.org/content/10.1101/2020.12.22.423931v1.full so the authors could not be aware of this work; however, it would be nice if the authors could comment on this work.

  4. Version published to 10.1101/515627v4 on bioRxiv
  5. Version published to 10.1101/515627v3 on bioRxiv
  6. Version published to 10.1101/515627v2 on bioRxiv
  7. Version published to 10.1101/515627v1 on bioRxiv