Editorial

Image-Based Systems Biology Marc Thilo Figge,1,2* Robert F. Murphy3,4*

SYSTEMS biology arose from the realization that organisms exhibit the properties of complex systems, in which behaviors of the whole cannot be predicted from analysis of individual components. This led to the demand for models that capture the complex relationships between components and how they give rise to observable behaviors at levels ranging from the subcellular to the ecological. Over the past twenty years, systems biology has largely been focused on acquisition of “omic” scale measurements, development of modeling approaches relevant to this scale, and creation and testing of models for particular systems. Most of this effort has been focused on reconstructing genetic regulatory networks and biochemical or metabolic pathways. Even though microscopy (especially fluorescence microscopy) has become an important tool in high throughput systems approaches, the spatial aspects of networks have been rarely studied. In those cases where they are, it is usually through compartmental models that treat the organization of cells, tissues, and organisms using conceptual subdivisions rather than through statistically and geometrically accurate representations that allow the interplay between networks and spatiotemporal organization to be captured. This Cytometry Part A Special Issue focuses on the growing discipline of Image-based Systems Biology that seeks to take full advantage of the information in images and establishes an essential connective link between experimental and theoretical examination of biological processes at a spatiotemporal level. The discipline generally combines three elements: (i) Acquisition and automated analysis of image data for high-content and high-throughput experimentation (1,2), 1

Applied Systems Biology, HKI-Center for Systems Biology of Infection, Leibniz-Institute for Natural Product Research and Infection Biology – Hans-Kn€oll-Institute (HKI), Jena, Germany

2

Faculty of Biology and Pharmacy, Friedrich Schiller University, Jena, Germany

3

Departments of Computational Biology, Biological Sciences, Biomedical Engineering and Machine Learning, Carnegie Mellon University, Pittsburgh, Pennsylvania

4

Freiburg Institute for Advanced Studies and Faculty of Biology, Albert Ludwig University of Freiburg, Germany

Cytometry Part A  87A: 459 461, 2015

(ii) Quantitative description of biological components (3,4), (iii) Construction of image-derived spatiotemporal models (5,6). The Special Issue includes twelve articles—one review and eleven research articles—and the word-cloud generated from all abstracts gives an intuitively accessible view of the questions and methods of the field today (see Fig. 1). The review by Medyukhina et al. (this issue, page 462 ) provides a detailed overview on Image-based Systems Biology with special emphasis on its application in the context of infections. Three steps of this approach—imaging, quantitative characterization, and modeling—are discussed and the application of these steps in the context of studying infection processes are considered. After summarizing the most relevant microscopy and image analysis approaches, ways to quantify infection processes are dicussed, and a number of different modeling techniques that exploit image-derived data to simulate host–pathogen interactions in silico are addressed. The study by Lockley et al. (this issue, page 471) convincingly demonstrates that selection between different mathematical models and identifiability analysis of model parameters, as commonly applied to temporal dynamics problems in systems biology, can as well be a powerful tool when extended to spatiotemporal imaging data. This was investigated in the context of reaction-diffusion models for fluorescence intensities measured in Dictyostelium amoebae that reorient in response to alternating gradients of mechanical shear flow. Herberg et al. (this issue, page 481) also present an approach which combines live-cell imaging, quantitative *Correspondence to: Marc Thilo Figge, Applied Systems Biology, HKI-Center for Systems Biology of Infection, Leibniz-Institute for Natural Product Research and Infection Biology – Hans-Kn€ oll-Institute (HKI), Jena, Germany. E-mail: [email protected] (or) Robert F. Murphy, Freiburg Institute for Advanced Studies and Faculty of Biology, Albert Ludwig University of Freiburg, Germany. E-mail: [email protected] Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/cyto.a.22663 C 2015 International Society for Advancement of Cytometry V

Editorial

Figure 1. Word-cloud created from the abstracts of all articles in this special issue.

image analysis, and multiscale modeling to study the growth of embryonic stem cells (ESC) giving rise to spatial structures that are vital for the maintenance of pluripotency. Quantitative measures of morphology and of spatial clustering were developed and applied to show that most ESC colonies contained only one cluster of cells with high self-renewing capacity and that these cells appeared preferentially located in the interior of a colony structure. Combining image analysis at the cellular level with mathematical modeling at the intracellular level revealed transcription factors responsible for the emergence of the observed patterns that could not have been derived from images directly. The stem cell theme is continued in the study by Jiang et al. (this issue, page 491) of adult human mesenchymal stem cell development. It focuses on the aggregation of these cells, in particular on the formation of three-dimensional (3D) cellular spheroids known to support expression maintenance of stemness marker genes in the cells. To this end, a scheme of image processing techniques was proposed consisting of a hybrid thresholding technique for efficient segmentation of cell clusters and a cell tracking method based on pairmatching with topological constraints. Mokhtari et al. (this issue, page 503) combined experiment and theory to study cell colocalization in murine bone marrow. Confocal fluorescence microscopy images of histological sections from murine bone marrow were generated and subsequently analyzed in an automated fashion. The quantitative analysis was combined with computer simulations of the experimental system for hypothesis testing with regard to the observed spatial colocalization of cells in the bone marrow. This analysis suggests that the plasma cell bone marrow survival niche facilitates colocalization of plasma cells with stromal cells and eosinophils, respectively, promoting plasma cell longevity. Segmentation of components is a key challenge in biological image analysis, and de Santos-Sierra et al. (this issue, page 513) introduce a graph-based unsupervised segmentation algorithm for structure characterization and modeling of cul460

tured neuronal networks. Their analysis was based on large scale phase-contrast images taken at high resolution through the life of a cultured neuronal network. The algorithm had a very low computational cost, scaling linearly with the image size, and processing automatically retrieved the whole network structure and all relevant morphological information characterizing neurons and neurites. Morevoer, their non-invasive phase-contrast images enable longitudinal analysis during the maturation of a single culture and the self-organization of the ensemble of neurons into a complex network. Computational approaches for automatic analysis of image-based high-throughput and high-content screens are investigated by Harder et al. (this issue, page 524). Their approach consists of cell segmentation, tracking, feature extraction, classification, and model-based error correction. It can be used for experimental optimization by extracting quantitative information which allows experimentalists to optimally choose and to verify the experimental parameters. The method was applied to a large-scale neuroblastoma screen, including the detection of rare division events such as multipolar divisions. An extensive quantitative evaluation was performed to separately evaluate the accuracies for segmentation, tracking, and classification. Cheung et al. (this issue, page 541) present another study that focuses on high-throughput screening, in particular, on image-based cell-resolved screening in a parallel microfluidic cytometer (PMC). This device is based on a one-dimensional scanning detector, a parallel array of flow channels, and new multiparameter analysis algorithms that operate on low-pixelcount images. Exploring a series of image-based live- and fixed-cell screening assays, a multiparametric linear weighted classifier was developed. It was concluded that the PMC would have the throughput and statistical power to permit a new capability for image-based high-sample-number pharmaceutical screening with suspension samples. The contribution by Wiliem et al. (this issue, page 549) describes a “bag of cells” approach for image classification that was applied in the context of identifying connective tissue

Editorial diseases by antinuclear antidbodies on HEp-2 cells. Applying image-based computer aided diagnosis, a “bag of visual words” approach was used to solve the image classification task, where each specimen was considered as a visual document containing visual vocabularies that are represented by its cells and that are used in the comparative classification of cells. This method offered a practical approach which could be deployed in high-throughput laboratory settings and could be generally applicable where examining the patterns of cells is required to make an inference on the specimen. Continuing the subcellular theme, the study by Steiniger III et al. (this issue, page 558) was motivated by the general limitation in microscopy that only a small number of biomarkers can typically be monitored simultaneously. A framework is described for comparing phenotypic cell states across biomarkers. This approach overcomes the current limitation of microscopy by not requiring costaining biomarkers on the same cells; instead, it is required that staining of biomarkers (possibly separately) on a common collection of phenotypically diverse cell lines is performed. The study suggests that many biomarkers provide redundant information about heterogeneity, such that the approach provides a practical guide for selecting independently informative biomarkers and, more generally, will yield insights into both the connectivity of biological networks and the complexity of the state space of biological systems. The topic of colocalization and its statistical analysis at the level of molecules is addressed in a study by Lagache et al. (this issue, page 568). Obviously, a rigorous quantitative analysis of molecule interactions in bioimaging of different populations is key for understanding the molecular orchestration of cellular processes. The authors present an overview of different colocalization methods followed by a quantitative comparison of their relative merits in different types of biological applications and contexts. It is found, using synthetic as well as biological image data, that object-based colocalization

Cytometry Part A  87A: 459 461, 2015

methods are statistically more robust than pixel-based methods, and allow for accurate quantification of the number of colocalized molecules. Concluding at the molecular level, the work by Wallrabe et al. (this issue, page 580) performs a quantitative analysis of protein interactions in the nucleation of actin filaments in live cells. A novel application of live cell 3-color fluorescence resonance energy transfer (FRET) microscopy is demonstrated, extending traditional analysis beyond energy transfer efficiency calculations. The experiments on cells from the MDCK cell line in combination with corresponding in vitro biochemical reconstitution for the same proteins were conducted to evaluate actin filament nucleation. This work emphasizes the power of 3-color FRET as a systems biology strategy for simultaneous evaluation of multiple interacting proteins in individual live cells. We hope that the assembly of these articles gives the reader a picture of the current directions in which imaging and image analysis are helping to shape systems biology. Just as systems biology must incorporate information about spatiotemporal organization, biological image analysis must also adapt to a more systems approach in which the outcome of image analysis is verifiable models (7).

LITERATURE CITED 1. Eliceiri KW, Berthold MR, Goldberg IG, Ibanez L, Manjunath BS, Martone ME, Murphy RF, Peng H, Plant AL, Roysam B, et al. Biological imaging software tools. Nat Methods 2012;9:697–710. 2. Pepperkok R, Ellenberg J. High-throughput fluorescence microscopy for systems biology. Nat Rev Mol Cell Biol 2006;7:690–696. 3. Murphy RF. Communicating subcellular distributions. Cytometry PART A 2010;77A: 686–692. 4. Mech F, Wilson D, Lehnert T, Hube B, Figge MT. Epithelial invasion outcompetes hypha development during candida albicans infection as revealed by an image-based systems biology approach. Cytometry Part A 2014;85A:126–139. 5. Zhao T, Murphy RF. Automated learning of generative models for subcellular location: Building blocks for systems biology. Cytometry Part A 2007;71A:978–990. 6. Figge MT, Garin A, Gunzer M, Kosco-Vilbois M, Toellner K-M, Meyer-Hermann M. Deriving a germinal center lymphocyte migration model from two-photon data. J Exp Med 2008;205:3019–3029. 7. Murphy RF. A new era in bioimage informatics. Bioinformatics 2014;30:1353.

461

Image-based systems biology.

Image-based systems biology. - PDF Download Free
163KB Sizes 3 Downloads 14 Views