Gene REgulatory Network Decoding Evaluations tooL (GRENDEL)

Category Cross-Omics>Pathway Analysis/Gene Regulatory Networks/Tools

Abstract GRENDEL is an open and extensible software toolkit that generates random ‘gene regulatory networks’ according to user-defined constraints on the network topology and kinetics.

It then simulates the state of each ‘regulatory network’ under various user- defined conditions (the experimental design) and produces simulated ‘gene expression’ data, including ‘experimental noise’ at a user defined level.

According to the manufacturer GRENDEL provides greater flexibility and realism than previously published ‘synthetic benchmarks’.

GRENDEL's more realistic network topologies Not only lead to lower accuracy estimates for all algorithms tested, but also they change estimates of which algorithms are more accurate under different ‘experimental designs’.

The manufacturers believe that GRENDEL will be useful both to experimentalists designing ‘gene expression’ studies and algorithm developers implementing and testing new computational approaches.

The manufacturers hope that, through both of these avenues, GRENDEL will help to advance the useful application of algorithms for reconstructions of ‘gene regulatory networks’.

The GRENDEL network process --

The artificial networks generated by GRENDEL are continuous-time ‘dynamical systems’ with three (3) independent types of ‘molecular species’: mRNAs, proteins and ‘environmental stimuli’ (e.g. extra-cellular glucose or iron).

In real networks, the relationship between a gene's mRNA and protein concentrations has been shown to be crucial for determining biologically relevant dynamics, as in certain oscillators.

Environmental stimuli, or signals, were included for the purpose of supporting time courses.

Signals are different than mRNAs and proteins in that they are driven by ‘external rules’ and are independent of the concentrations of mRNAs and proteins.

Computationally generating random ‘biological networks’ involves two (2) modular steps: topology generation and kinetic parameterization.

The ‘topology generation’ step defines the reagents, catalysts and products of each reaction.

After generating a graph indicating which genes regulate which other genes, GRENDEL ‘chooses parameters’ for the differential equations that determine the concentration of each protein and each mRNA.

These parameters allow for the simulation of both a network's responses to ‘environmental changes’ and the effects of ‘genetic interventions’ on those responses.

After generating a network, GRENDEL exports it in Systems Biology Markup Language (SBML).

Networks specified in SBML can be simulated by using one of several SBML integration programs, including COPASI (see G6G Abstract Number 20296), CellDesigner (see G6G Abstract Number 20159) and SBML ordinary differential equation (ODE) Solver Library (SOSlib).

The manufacturer's software uses SOSlib (see below...) to deterministically integrate the ODEs that define the dynamical system, resulting in ‘noiseless expression’ data.

Simulated ‘experimental noise’ is then added to the data according to a log-normal distribution, with user-defined variance.

The networks that the manufacturer's method produces could be simulated with ‘biological noise’ by using an SBML-based stochastic integrator, such as, Dizzy -

(Dizzy is a software tool for stochastically and deterministically modeling the ‘spatially homogeneous kinetics’ of integrated large-scale genetic, metabolic, and signaling networks.

Notable features include a modular simulation framework, reusable modeling elements, complex kinetic rate laws, multi-step reaction processes, ‘steady-state noise estimation’, and spatial compartmentalization).

SOSlib (SBML ODE Solver Library) -

SOSlib is both a programming library (API) and a set of command-line applications for symbolic and numerical analysis of a system of ordinary differential equations (ODEs), derived from a (bio)-chemical reaction network encoded in SBML.

Benefits of using ‘simulated networks’ such as GRENDEL --

One of the benefits of using simulated networks to evaluate reconstruction algorithms is the statistical power one gets from being able to generate many networks sampled from the same distribution.

If an algorithm performs very poorly at reconstructing a specific subset of networks, the ability to generate large populations of networks enables developers to identify the weaknesses of their method.

In silico benchmarks also allow for properties of ‘regulatory networks’, such as degree distributions, experimental noise, biological noise and network size, to be varied independently of one another. This helps to identify the properties that contribute most to ‘reconstruction error’.

Simulated networks also have great potential as cost effective tools for determining the ‘optimal experimental design’ to use with a given ‘network reconstruction method’.

The results obtained with ‘simulated networks’ are only a first step in evaluation that must ultimately be followed by application to ‘real biological systems’.

At present, simulated networks are rough approximations that omit many important aspects of ‘biological systems’, including localization and post- translation modifications.

System Requirements

Contact manufacturer.


Manufacturer Web Site GRENDEL

Price Contact manufacturer.

G6G Abstract Number 20585

G6G Manufacturer Number 104188