April 26, 2017, Wednesday

From BIMIB

Jump to: navigation, search

Retronet

The analysis of biological systems relies more and more on computational and mathematical methods. The goals of such analysis are multifarious; among the most important ones is the discovery of the biochemical and genetic machinery responsible for pathology development, its control and, possibly, elimination. Such discoveries also rely on an understanding of the spatio-temporal development of biological phenomena, their cause (often "mutations") and their effects on different scales.

The RetroNet project intends to address this problem and others by i) sharing data and knowledge needed for a new integrative research approach in medicine, ii) sharing or jointly develop multiscale models, simulators and analysis tools, with particular attention to the development of Colon Rectal Cancer (CRC) and some of its metastatic effects, and, iii) creating the prototype of a collaborative environment supporting research in this highly interdisciplinary field, by leveraging the experience matured from of previous FP6 experiences.

The RetroNet project will concentrate on the development and tuning of algorithms for detecting of emerging behavior from cells ensembles, by searching, analysing and formulating hypotheses of various feedback cycles in biological systems. The approach will leverage several Control-Theoretic concepts; especially the notions of state-estimation and control-policy learning as implicit drivers of biological behavior selection. The emerging-behavior detection algorithms will consider the content of Pathway and Models Databases and knowledge directly gained from clinicians and biologists running bio-banks or wet-laboratory focussed research.


Contents

GESTODifferent

A Cytoscape plugin for the identification of Boolean Gene Regulatory Networks  describing the stochastic differentiation process

Threshold-dependent attractor transition network and the tree-like TES landscape. The circle nodes are attractors of an example NRBN, the edges represent the relative frequency of transitions from one attractor to another one, after a 1 time step-flip of a random node in a random state of the attractor (performed an elevated number of times). An arbitrary threshold is introduced to account for the differentiation degree and here we show three different values of the threshold, i.e.: 0, 0.15 and 1. TESs, i.e. strongly connected components in the threshold-dependent attractor transition network are represented through dotted lines and the relative threshold is indicated in the subscripted index. In the right diagram it is shown the tree-like representation of the TES landscape, which determines the differentiation tree for this NRBN.
Threshold-dependent attractor transition network and the tree-like TES landscape. The circle nodes are attractors of an example NRBN, the edges represent the relative frequency of transitions from one attractor to another one, after a 1 time step-flip of a random node in a random state of the attractor (performed an elevated number of times). An arbitrary threshold is introduced to account for the differentiation degree and here we show three different values of the threshold, i.e.: 0, 0.15 and 1. TESs, i.e. strongly connected components in the threshold-dependent attractor transition network are represented through dotted lines and the relative threshold is indicated in the subscripted index. In the right diagram it is shown the tree-like representation of the TES landscape, which determines the differentiation tree for this NRBN.

GESTODifferent is a plugin for Cytoscape, based on a dynamical model of cell differentiation originally described in "Villani M, Barbieri A, Serra R (2011) A Dynamical Model of Genetic Networks for Cell Differentiation. PLoS ONE 6(3): e17703. doi:10.1371/journal.pone.0017703".


The model associates the type and the specific degree of differentiation of a cell to the stability (i.e., resistance to noise and perturbations) of the steady states of its underlying Gene Regulatory Network (GRN), here modeled as a Noisy Random Boolean Network (NRBN). The steady states represent coherent gene activation patterns, driving the correct functioning of each distinct cell type. NRBNs reproduces key properties of the differentiation process, e.g. the existence of toti- and multi-potent stem cells, the presence of different degrees of differentiation and the phenonomenon of stochastic differentiation, according to which a population of multi- or toti-potent cells generate progenies of different types through a stochastic process.


GESTODifferent allows to select a desired lineage commitment tree (also defined as differentiation tree), e.g., that of hematopoietic cells or intestinal crypt cells, and to search for the NRBN displaying the correct emergent dynamical behaviour. Given certain structural features of the NRBN (e.g. the number of genes, the topology of the connections, the set of updating functions, etc.), GESTODifferent generates a number of randomly generated networks, whose dynamical behaviour will be matched against the expected differentiation tree. The resulting NRBNs can be successively displayed and analyzed through Cytoscape and Cytoscape plugins and can be used, for instance, as GRNs in multiscale models of complex biological systems.

Example screenshot of GESTODifferent plugin.
Example screenshot of GESTODifferent plugin.
Example workflow of GESTODifferent.
Example workflow of GESTODifferent.


In relation to the BIMIB research, GESTODifferent was used to perform a key part of the simulations of multiscale models of intestinal crypt dynamics, leading to the following publications:

  • Graudenzi, A., Caravagna, G., De Matteis, G., Mauri, G., Antoniotti, M. (2012): A multiscale model of intestinal crypts dynamics. In Proc. of the Italian Workshop on Artificial Life and Evolutionary Computation, WIVACE 2012; ISBN: 978-88-903581-2-8.
  • Graudenzi, A., De Matteis, G., Alhazov, A., Caravagna, G., Mauri, G., Antoniotti, M. (2011): Multiscale multicellular modeling for the description of intestinal crypt dynamics and the colorectal cancer development (long abstract and poster). Accepted for the publication on the proceedings of NETTAB 2011 Workshop focused on Clinical Bioinformatics. October 12-14, 2011, Pavia, Italy.


Manual

The manual of the plugin can be downloaded at this link.



License

The plugin is release under a BSD-like license which can be found in the COPYING file.

Performances

In the right table you can find the average value and the relative standard deviation of the computation times of the different tasks of the plugin, namely: a) the network creation time, b) the attractors' search time, c) the attractors' perturbations time, d) the time of definition of the differentiation tree, e) the time of the matching between the output and the input tree, and f) the total computation time. The unit of measure is seconds.  The data refers to 2500 different simulations of 100 nodes NRBNs, with scale-free topology, canalyzing functions only and average connectivity |A|=2. Notice that the high standard deviation hints at a high (and expected) dispersion of the computation times, due to the known general heterogeneity of this kind of dynamical system.

Computation times. Unit of measure: seconds
Computation times. Unit of measure: seconds


How to install

IMPORTANT NOTE: GESTODifferent is compatible with Cytoscape 2.8, it has not been tested with other version of the application.

  1. Download the file at this link.
  2. Place the jar file in the Plugin folder, e.g. ../Applications/Cytoscape_v2.8.2/plugins
  3. Restart Cytoscape.
  4. You will find the item GESTODifferent in the Plugin menu.
  5. Run the plugin.


Example differentiation trees

At the following links you can find the sif file of two example differentiation trees, a trivial one and the real differentiation tree of intestinal crypts (see the relative figures below):

Note. We remark that given the high complexity of the differentiation tree of intestinal crypts the computation time can accordingly be very large. 

Example differentiation tree. The descendent of a unique stem cell type are two distinct differentiated cell types. Schematic representation of the crypt differentiation tree. Stem cells are the root of the tree, while the 4 differentiated cell types (i.e. Paneth, Goblet, enteroendocrine, absorptive or enterocyte) are the leaves. TA stands for transit amplifying stage, which is known to be the intermediate state between stem and fully differentiated stages.

Left figure: Example differentiation tree. The descendent of a unique stem cell type are two distinct differentiated cell types. Right figure: Schematic representation of the crypt differentiation tree. Stem cells are the root of the tree, while the 4 differentiated cell types (i.e. Paneth, Goblet, enteroendocrine, absorptive or enterocyte) are the leaves. TA stands for transit amplifying stage, which is known to be the intermediate state between stem and fully differentiated stages.


Credits and contacts

GESTODifferent was developed by Silvia Crippa, under the supervision of Giulio Caravagna, Alex Graudenzi and Marco Antoniotti. For any question, comment, suggestion please write to: marco.antoniotti [a] disco.unimib.it.