A WEB-BASED REPOSITORY OF REPRODUCIBLE SIMULATION

EXPERIMENTS FOR SYSTEMS BIOLOGY

Michael A. Guravage and Roeland M. H. Merks

Centrum Wiskunde & Informatica (CWI)

Netherlands Consortium for Systems Biology & Netherlands Institute for Systems Biology (NCSB-NISB)

Science Park 123, 1098 XG Amsterdam, The Netherlands

Keywords:

SED-ML, SBML, MIASE, Simulation, Modeling, Plone, CMS.

Abstract:

Systems Biology requires increasingly complex simulation models. Effectively interpreting and building upon

previous simulation results is both difﬁcult and time consuming. Thus, simulation results often cannot be

reproduced exactly; making it difﬁcult for other modellers to validate results and take the next step in a

simulation study.

The Simulation Experiment Description Mark-up Language (SED-ML), a subset of the Minimum Information

About a Simulation Experiment (MIASE) guidelines, promises to solve this problem by prescribing the form

and content of the information required to reproduce simulation experiments. SED-ML is detailed enough to

enable automatic rerunning of simulation experiments.

Here, we present a web-based simulation-experiment repository that lets modellers develop SED-ML compli-

ant simulation-experiment descriptions The system encourages modellers to annotate their experiments with

text and images, experimental data and domain meta-information. These informal annotations aid organi-

sation and classiﬁcation of the simulations and provide rich search criteria. They complement SED-ML’s

formal precision to produce simulation-experiment descriptions that can be understood by both men and ma-

chines. The system combines both human-readable and formal machine-readable content, thus ensuring exact

reproducibility of the simulation results of a modelling study.

1 INTRODUCTION

Systems biologists unravel the inner workings of bi-

ological systems in a continuous cross-talk between

predictive computer simulation models, and biolog-

ical experiments to test the model predictions. To

build models efﬁciently and compare model predic-

tions with experiments, effectively interpreting sim-

ulation results and building upon previous work is

crucial. Unfortunately, it is both difﬁcult and time

consuming, because, 1) publishing exact model def-

initions and working simulation codes is not (yet)

common practice, and 2) the exact steps taken and

the information required for reproducing the simula-

tion (e.g. model parameters) are rarely recorded ex-

actly. As a result, trying to reproduce published sim-

ulation results can become a research project in itself.

This hampers progress in systems biology. The dif-

ﬁculty in validating results impedes the progress of

simulation studies.

A range of public databases exists for depositing

computational models in a systematic way, including

the BioModels database (Le Nov

ere et al., 2006), the

modelDB (Hines et al., 2004), the Database of Quan-

titative Cellular Signalling (DOQCS) (Sivakumaran

et al., 2003) and Physiome (Yu et al., 2009; Yu et al.,

2011). Although these databases make it much easier

to retrieve the deﬁnitions and codes of published sim-

ulation models, reproducing an observation in a sim-

ulation model also requires exact knowledge of many

additional factors including the initial conditions, the

parameters used, the integration times, etc. In other

words, the model databases contain the ingredients

for a simulation experiment, but the recipe is lacking.

The Simulation Experiment Description Markup

Language (SED-ML) (K

ohn and Le Nov

ere, 2008)

implements a subset of the Minimum Information

About a Simulation Experiment (MIASE) guidelines.

MIASE is an emerging standard that promises to

solve this problem by prescribing the form and con-

tent of the information required to reproduce simula-

tion experiments. SED-ML is a recipe template, an

instance of which include a description of and refer-

ence to existing, preferably curated, models, inputs,

134

A. Guravage M. and M. H. Merks R..

A WEB-BASED REPOSITORY OF REPRODUCIBLE SIMULATION EXPERIMENTS FOR SYSTEMS BIOLOGY.

DOI: 10.5220/0003598001340141

In Proceedings of 1st International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH-2011), pages

134-141

ISBN: 978-989-8425-78-2

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

procedures and outputs that, together, describe of how

to use the models in simulations that constitute an ex-

periment. While SED-ML is detailed enough to en-

able automatic rerunning of simulation experiments,

its XML-based format makes it difﬁcult for modellers

to interpret. Also, SED-ML focuses on computational

models of biological processes written in, e.g., SBML

format; whereas the simulations most difﬁcult to re-

produce are typically written in general-purpose lan-

guages. Our repository accommodates models writ-

ten in both XML and general-purpose languages like

C++ by allowing model references to include both

model databases and source code repositories.

SED-ML is an grammar for describing simulation

experiments. Its speciﬁcity allows one to describe a

simulation experiment in sufﬁcient detail to reproduce

the simulation results automatically. SED-ML even-

tually can act as a scripting language to describe the

steps leading to a published simulation result. While

SED-ML ensures that the simulation-experiment de-

scription is unambiguous, it is difﬁcult to interpret

the intention of the modeller by reading only his

SED-ML code. Combining complimentary informal

descriptive information with SED-ML’s formal spec-

iﬁcation ensures that simulation experiments descrip-

tions can be interpreted by both humans and ma-

chines. Adherents of Literate Programming (Knuth,

1984) have long argued that the complementary na-

ture of informal descriptions and formal speciﬁca-

tions facilitates the retention and transfer of informa-

tion.

For the modeller, creating simulation-experiment

descriptions in our repository is analogous to keep-

ing a laboratory notebook. Models can be described

in the context of real simulations at an appropriate

level of abstraction and detail, and the complementary

nature of informal and formal speciﬁcations encour-

ages correctness and completeness. As a result, sim-

ulation experiments are both more easily reproduced

and more readily understood. Using the repository

thus facilitates collaboration between modellers. The

repository can be tailored to realise any collaboration

strategy, e.g., a repository’s contents can be accessi-

ble to the public or restricted to an enumerated list of

individuals. Modelers can invite colleagues to col-

laborate on speciﬁc simulation-experiment descrip-

tions. Workﬂows shepherd simulation-experiment de-

scriptions through a series of states from inception

to completion; so a modeller can easily see which

simulation-experiment descriptions are in develop-

ment, which have been submitted for approval and

which have been tested and validated. Experimen-

tal collaborators can evaluate a repository of quali-

ﬁed simulation-experiment descriptions that describe

models in the context of a real simulation.

In this paper, we argue that placing simulation-

experiment descriptions in a repository that allows an-

notating them with prose descriptions, domain meta-

information and experimental data, and that guides

their progress through a publication workﬂow, will

signiﬁcantly improve the usability of the SED-ML

recipes. Our ﬁnal goal is to see published simula-

tion experiments refer to simulation-experiment de-

scriptions in our repository. Those descriptions will

contain both machine-executable and human-readable

recipes that allow modellers to interpret, evaluate and

reproduce the simulations and build upon the pub-

lished results.

2 THE MODEL REPOSITORY

2.1 Requirements

We aimed to construct a system that allows mod-

ellers to create simulation-experiment descriptions

conforming to the MIASE guidelines, according to

the following system requirements:

2.1.1 Ease of Use, Encourage Annotation

The system should help modellers along the process

of creating and managing their simulation-experiment

in ways that are natural and easy for them. The top

level of a simulation-experiment description hierar-

chy is comprised of ﬁve classes:

• Models

• Simulations

• Tasks

• Outputs

• Data Generators

The system assists the modeller in creating the

various elements in the correct order and with valid

data. In addition to the minimum data prescribed by

the MIASE guidelines, the system should encourage

the modeller to annotate each part of a simulation-

experiment description with descriptive information

at an appropriate level of abstraction. The editing

environment should be familiar to anyone who has

worked with ofﬁce suite software. The system should

also allow the user to include experimental data such

as images and animations, and generic and domain-

speciﬁc meta-information. This data should be in-

dexed automatically and be made available as search

criteria.

A WEB-BASED REPOSITORY OF REPRODUCIBLE SIMULATION EXPERIMENTS FOR SYSTEMS BIOLOGY

135

2.1.2 Data Security

The system should have several levels of security,

such that users can choose to keep a project from

themselves, share it with coworkers or reviewers,

or make it public after the results have been pub-

lished in a scientiﬁc journal. Modelers own their

simulation-experiment description, and they cannot

be viewed or altered without their owner’s consent.

The owner of a simulation-experiment description is

able to invite fellow modellers to collaborate on spe-

ciﬁc simulation-experiment descriptions. Thus, the

system should make a distinction between anonymous

and authenticated users. Permissions based on roles,

e.g., anonymous visitor, authenticated member, mod-

eller and curator can be used to delegate responsibili-

ties and precisely control access and visibility.

2.1.3 Data Accessibility

Someone using the system should easily be able

to search through the collection of simulation-

experiment descriptions using a variety of search cri-

teria. Finding the the simulation-experiment descrip-

tion he or she wanted, a modeller should be able

to browse through it and, if appropriate, to export

the entire simulation-experiment description to a ﬁle

conforming to the SED-ML standard. Over time,

simulation-experiment descriptions should resemble

‘laboratory notebooks’ that contain their entire his-

tories from inception through development to publi-

cation. Additional tools should make it possible to

reproduce automatically simulations based on the ex-

ported SED-ML ﬁles.

2.2 Architecture

Based on these system requirements, we implemented

a prototype of the system based on the content man-

agement system Plone

(Pastore, 2006). We choose

Plone because its concept of ‘roles’ clariﬁes responsi-

bilities: this allows modellers to concentrate on creat-

ing their content. Plone’s development tools allowed

us to turn Uniﬁed Modeling Language (UML)

mod-

els into running Plone code.

Plone is a free and open source content man-

agement system written in Python

and built on top

of the Zope

application server. It stores all its

information in Zope’s built-in transactional object

database (ZODB) (Fulton, 2000). We have developed

www.plone.org

www.omg.org/spec/UML/2.1.2

www.python.org

www.zope.org

our simulation-experiment description repository by

modelling a set of SED-ML content types in UML.

Our models are input to ArchGenXML (Auersperg

et al., 2007) - a code generating tool will turn our

UML models into a Plone add-on product that we in-

stall in a Plone portal running on the Zope server.

2.2.1 Performance

The number of SEDs a repository can hold is limited

only by the physical capacity of the server on which

repository runs.

The Oscillation to Chaos (O2C) SED exam-

ple (K

ohn and Le Nov

ere, 2008) is a hierarchy of 18

SED objects in our repository that occupy 30K in the

Zope database. O2C’s exported SED-ML representa-

tion, stripped of all its annotations, is a mere 3.5k.

To evaluate the effect of the Varnish cache we used

ApacheBench

to collect median performance data

from both the default Apache/Varnish/Plone route,

and directly from Plone via the zeo-clients. Figure 1

shows how efﬁcacious a caching server becomes as

the number of concurrent requests increase.

Figure 1: Median Total Request Time.

2.2.2 Deployment

The SED repository can be installed on any Zope ap-

plication server running Plone 3. We deploy Plone

on a shared virtual server with two quad-core Xeon

E5405 processors and 4GB of RAM.

Plone is conﬁgured in a standard production en-

vironment (Aspeli, 2007) with an Apache web server

acting as a reverse proxy for a Varnish caching server

that balances requests over ﬁve zeo clients connected

to a single Zope application server.

Our memory footprint is less than 100MB for the

Zope application server and an average of 300MB for

each zeo-client.

httpd.apache.org/docs/2.0/programs/ab.html

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

136

2.3 Object Model

As a formal representation of the simulation ex-

periments, we made use of the Simulation Exper-

iment Description Object Model (SED-OM). The

SED-OM is a formal representation of the MIASE

guidelines expressed using the Uniﬁed Modeling

Language (UML). From the object model we derive

the Simulation Experiment Description Meta Lan-

guage (SED-ML) schema that represents the grammar

for SED-ML compliant documents. The SED-ML

XML schema is distributed in XSD

format - a W3C

consortium standard for deﬁning an XML Document.

Beginning with the SED-OM schema we reverse-

engineered our own object-model - annotated with

Plone supplied tagged-values that specify how data

members look and behave in the Plone UI. For this

purpose, we chose to use an XML editor Named Oxy-

gen

and a UML editor named Poseidon

. Our repos-

itory and its custom content types are based on the

Level 1 Version 1 (Draft) version of the SED-ML

schema (Bergmann, 2010).

Casting the SED-ML XML schema as a set of

custom content types that represent the objects in the

SED-OM allows nontechnical users to use the content

types as building blocks to create and manage Simu-

lation and Experiment Descriptions. Modelers are not

confronted with complex XML syntax, but create and

manage simulation-experiment descriptions through a

set of clear and concise forms. Figure 2 shows the

Simulation edit form. Its data ﬁelds conform to the

SED-ML XML schema simulation element. The de-

tailed description ﬁeld uses Plone’s own powerful rich

editor with text formatting and image and link inser-

tion abilities. A simulation-experiment description’s

data are stored in such a way that it can be exported

in SED-ML compliant XML.

2.3.1 Code Generation

To generate Plone add-on modules from the UML

class model, we used ArchGenXML

. Arch-

GenXML is a code-generator for Plone. Archetypes

is a framework for building content objects based on

schemas and provides features such as automatically

generating editing and presentation views, assigning

unique content Ids, registering content types with

various Plone management tools and storage trans-

parency. Our newly generated content types behave

www.w3.org/XML/Schema

www.w3c.org

www.oxygenxml.com

www.gentleware.com

plone.org/products/archgenxml

Figure 2: Simulation Edit Form.

like native content types in that they adhere to Plone’s

document workﬂows and security model.

ArchGenXML automatically generates the edit-

ing and presentation views for each new content type.

In the UML model, we annotate class member data

with tagged-values that describe how it should look

and behave. For example, the standard stipulates that

the top-level simulation-experiment description class

must have an integer version number. The tagged-

values assigned to it indicate that it is a required ﬁeld,

that it must be an integer, what its label and descrip-

tion text are and its default value.

Tagging simulation-experiment descriptions with

standard and domain meta-information will facilitate

their organisation and classiﬁcation while simultane-

ously providing rich search criteria. Plone imple-

ments the Dublin Core (Weibel et al., 1998) meta-

data - a small set of text elements through which

Plone content types can be described and catalogued.

Their values are indexed in Plone’s internal cata-

logue and can be used in searches. In addition to

the Dublin Core stipulated ’description’ element, each

simulation-experiment description content type has a

detailed description ﬁeld in the form of a rich text ed-

itor, that the modeller can use for writing prose de-

scriptions of the simulation experiments.

The simulation-experiment description object

model speciﬁc two vocabularies: a model’s encod-

A WEB-BASED REPOSITORY OF REPRODUCIBLE SIMULATION EXPERIMENTS FOR SYSTEMS BIOLOGY

137

ing formalism and a simulation’s algorithm desig-

nation. The former is a simple list including e.g.,

SBML, CellML and C++. The latter describes the

Kinetic Simulation Algorithm Ontology (KiSAO)

encoded in the vocabulary description markup lan-

guage (VDEX)

. Both vocabularies, while used by

the repository, are managed separately, which allows

the site administrators to modify the contents of the

vocabularies without adversely effecting the existing

models and simulations that use them.

2.3.2 Collaboration Strategies

Plone’s workﬂows control the progress a modelling

project through a series of states from inception to

publication. Figure 3 shows the states and transi-

tions in a restricted Plone workﬂow. By customis-

ing these workﬂows, we can realise a variety of col-

laboration strategies. For example, the current work-

ﬂow for our several simulation-experiment descrip-

tion repositories makes them accessible only to mem-

bers of their respective modelling groups. The work-

ﬂow progressively expands the audience as individual

simulation-experiment descriptions proceed through

a series of approvals, ﬁrst to other members of the re-

search group, then to portal members, and ﬁnally to

the general public.

draf t

start

private

pending

external

internal

show

hide

submit

publish-internally

publish-externally

retract/reject

publish-internally

retract/reject

publish-externally

Figure 3: A Restricted Plone Workﬂow.

2.3.3 Exporting SED-ML Compliant XML

In order to use the repository for automatically re-

producing simulation experiments, the system must

have facilities to export the simulation-experiment

description to SED-ML compliant XML. We have

www.ebi.ac.uk/compneur-srv/kisao

www.imsglobal.org/vdex

achieved this by customising the code generated by

ArchGenXML.

The simulation-experiment description class is the

root of a hierarchy that, taken in its entirety, rep-

resents a complete simulation-experiment descrip-

tion. We added custom python code to each class

in this hierarchy to enable their instances to create

an SED-ML representation of themselves and request

their children to do likewise. The code was writ-

ten with Python’s light-weight implementation of the

Document Object Model interface. A document tab

named ’Export this’ appears at the bottom of ev-

ery simulation-experiment description. It is linked

to a browser-view - a special class that links URLs

with speciﬁc bits of code in Zope’s component ar-

chitecture. The tab, when clicked, sets a simulation-

experiment description’s export process in motion.

Concluding, we have developed a set of custom

content types that faithfully represent the SED-ML

XML schema, which can be installed into a Plone por-

tal running on a Zope application server customised to

local needs.

2.4 Usage

2.4.1 Data Entry

The user interacts with our system via a Plone-based

web interface. Each SED-ML content type has its

own presentation and editing forms. Creating a

simulation-experiment description consists of navi-

gating to the repository and clicking on the ‘Add

SED’ menu button. A new simulation-experiment de-

scription edit form appears and solicits the required

information, and an appropriate amount of descrip-

tive information. As speciﬁed in the schema, the

SED-ML name space and level ﬁelds have default val-

ues. Since the simulation-experiment description ob-

ject is the root of a new simulation experiment hierar-

chy, after clicking the simulation-experiment descrip-

tion’s ‘Save’ button, the add items menu presents a

new list of content items the modeller can insert at this

point. This process continues until the simulation-

experiment description is complete.

2.4.2 Searching

Since we speciﬁed which information would be in-

dexed while we modelled our SED-ML content types

in UML, searching through the repository is now a

straightforward matter of typing our search criteria

into the search box.

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

138

2.4.3 Reproducing Simulation Experiments

When presented with a complete simulation-

experiment description, especially one written in a

general purpose programming language and not (yet)

reproducible automatically, the modeller can simply

retrieve the recipe from the repository. Section 3 is

a case study describing how to build a simulation-

experiment description whose model is written in

C++.

If a modeller wishes to reproduce the results of a

simulation-experiment description whose models are

written in a dialect of XML, e.g., SBML, and whose

simulations are, e.g., uniform time course, he or she

can export the simulation-experiment description in

SED-ML compliant XML directly by clicking on the

simulation-experiment description’s ‘Export this’ tab.

The repository will return a ﬁle whose name is derived

from the simulation-experiment description’s Id. The

modeller can use this ﬁle as input to one or another

simulation system such as COPASI (Hoops et al.,

2006).

2.4.4 In Conclusion

The repository is a resource with which a modeller

can fully describe a simulation-experiment descrip-

tion recipe. The SED-ML compliant content types

assist in this by soliciting and validating information

as it is entered. Content and meta-data are indexed

automatically and available as search criteria. A com-

pleted simulation-experiment description should con-

tain the descriptive information to allow the simula-

tion experiment to be repeated manually, and if appro-

priate its contents can be exported in SED-ML com-

pliant XML to be rerun automatically.

3 EXAMPLE SED COMPOSITION

To demonstrate how to compose a SED, we have cho-

sen an experiment that uses the VirtualLeaf (Merks

et al., 2011), an open-source framework for cell-

based modelling of plant tissue growth and develop-

ment. This particular experiment shows how auxins,

a family of plant growth hormones, can accumulate in

growing plant tissue to form bulbous patterns.

The VirtualLeaf distribution contains the Virtual-

Leaf framework software, various dynamical models

and sample simulation data; so acquiring, installing

and running the experiment is simple and comprises

only a few steps.

1. Download the VirtualLeaf Framework source

bundle from its Google Code repository.

2. Compile and install the VirtualLeaf software.

3. Choose a dynamical model to use, i.e. auxin

growth.

4. Choose the simulation data describing the experi-

ments initial conditions.

5. Run the experiment.

6. Aggregate the resulting experiment snapshots into

a single animation.

In a simulation experiment description the informa-

tion in these various steps is distributed over several

SED-ML elements as follows.

• The SED contains two Model elements. The ﬁrst

model describes the VirtualLeaf framework. The

second model describes the auxin growth model

plug-in that describes the dynamic character of

growing cells. Both models contain unique identi-

ﬁers naming the particular version of the Virtual-

Leaf software used in the simulation experiment

and its location in a googlecode repository from

where the source code can be downloaded.

• A single Simulation element contains a leaf de-

scription ﬁle that describes the initial conﬁgura-

tion of cells and their various properties; it also

contains a KiSAO designation of the type of algo-

rithm used.

• A single Task element associates the models with

the simulation as a single experiment.

• A single DataGenerator element describes how

to aggregate the experiment’s results, in this case

screen shots, into an animation; it also includes a

shell script that performs the aggregation.

• A single Output element describes the experi-

ments results, and points to a the dataGenerator

that tell how to reproduce them.

Figure 4 shows the repository’s output element

view of the animation generated by concatenating the

VirtualLeaf’s multi-screenshot output.

A screencast demonstrating how to build these el-

ements is available on our Repository’s public web-

site: sed.project.cwi.nl.

3.1 Run Simulations Automatically

Currently, only XML-based, e.g., SBML, simulations

can be reproduced automatically; using tools like CO-

PASI. For the moment, our SEDs whose models are

written in general-purpose languages contain descrip-

tive information in each SED element to help the

reader understand the simulation and then run it man-

ually.

A WEB-BASED REPOSITORY OF REPRODUCIBLE SIMULATION EXPERIMENTS FOR SYSTEMS BIOLOGY

139

Future versions of the repository will read a SED’s

resulting SED-ML ﬁle and generate scripts to repro-

duce the simulation (semi-) automatically using an

approach similar to Python’s buildout (Aspeli, 2007)

system where conﬁguration recipes preform common

tasks such as downloading, compiling and installing

a software source distribution or executing a program

together with its inputs.

4 DISCUSSION

In this paper, we have described a web-based

repository that assists modellers in creating

simulation-experiment descriptions that conform

to the Simulation Experiment Description Markup

Figure 4: Repository Output Element View of a VirtualLeaf

Animation.

Language (SED-ML) - a format for the imple-

mentation of the MIASE guidelines. Modelers

can build simulation-experiment descriptions in an

intuitive way, and annotate them with descriptions,

experimental data and domain meta-information.

This information is indexed automatically and made

available as search criteria. Finally, the contents of

simulation-experiment descriptions can be exported

in SED-ML compliant XML.

Our web-based simulation experiment repository

allows modellers to create and manage simulation-

experiment descriptions in a way analogous to keep-

ing a laboratory notebook. Models can be described

in the context of real simulations at an appropriate

level of abstraction and detail. The entire prove-

nance of a simulation experiment can be preserved

and retraced. Ideally, published simulation experi-

ment will refer to simulation-experiment descriptions

in our repository that are machine-executable and

human-readable recipes that allow modellers to repro-

duce simulations and build upon the published results.

The beneﬁt to the modellers that they have at

their disposal a repository of qualiﬁed simulation-

experiment descriptions that describe models in the

context of real simulations. Their detailed technical

speciﬁcation and associated descriptive information

will help reduce the time and effort needed to repro-

duce simulation results, and provide an excellent start

for reusing and extending models.

Maintaining a proper laboratory notebook re-

quires a great deal of discipline. Adequately anno-

tating a simulation-experiment description is equally

difﬁcult, if not more. For while a laboratory note-

book may be a private work record, it is our inten-

tion that once a simulation study has been published

its simulation-experiment descriptions in our repos-

itory will be read and used by many. The danger

for simulation-experiment descriptions, like any doc-

umentation, is that they will be left incomplete.

While our repository can export SED-ML com-

pliant XML, it cannot (yet) convert it back into a

SED-ML object hierarchy. We are contemplating

writing an import facility to read SED-ML ﬁles into

our repository. We would also like to update our

Simulation Experiment Description Object Model to

conform to the SED-ML Level 1 Version 1 (Final).

Currently, only XML-based, e.g., SBML, simulations

can be reproduced automatically; future versions will

also implement scripts that (semi-)automatically re-

produce simulations implemented in general-purpose

languages, e.g., using an approach similar to Python’s

buildout system.

We hope that modellers will be motivated to write

reproducible simulation-experiment descriptions us-

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

140

ing our repository when they experience the beneﬁts

of having full and immediate on-line access to their

work.

ACKNOWLEDGEMENTS

This work was (co)ﬁnanced by the Netherlands Con-

sortium for Systems Biology (NCSB) which is part

of the Netherlands Genomics Initiative / Netherlands

Organisation for Scientiﬁc Research.

We would also like to thank our colleagues in the

NCSB-NISB Biomodeling and Biosystems Analysis -

Life Sciences Group at CWI for their cooperation in

developing and testing the repository.

REFERENCES

Aspeli, M. (2007). Professional Plone Development. Packt.

Auersperg, P., Klein, J., and van Rees, R. (2007). Arch-

GenXML code generator.

Bergmann, F. (2010). Simulation Experiment Descrip-

tion Markup Language (SED-ML): Level 1 Version

1 (Draft).

Fulton, J. (2000). Introduction to the Zope Object Database.

In Proceedings of the 8th International Python Con-

ference.

Hines, M., Morse, T., Migliore, M., Carnevale, N., and

Shepherd, G. (2004). ModelDB: a database to sup-

port computational neuroscience. Journal of Compu-

tational Neuroscience, 17(1):7–11.

Hoops, S., Sahle, S., Gauges, R., Lee, C., Pahle, J., Simus,

N., Singhal, M., Xu, L., Mendes, P., and Kummer,

U. (2006). COPASI–a COmplex PAthway SImulator.

Bioinformatics, 22(24):3067.

Knuth, D. (1984). Literate programming. The Computer

Journal, 27(2):97.

ohn, D. and Le Nov

ere, N. (2008). SED-ML–An XML

Format for the Implementation of the MIASE Guide-

lines. In Computational Methods in Systems Biology,

pages 176–190. Springer.

Le Nov

ere, N., Bornstein, B., Broicher, A., Courtot, M.,

Donizelli, M., Dharuri, H., Li, L., Sauro, H., Schilstra,

M., Shapiro, B., et al. (2006). BioModels Database: a

free, centralized database of curated, published, quan-

titative kinetic models of biochemical and cellular sys-

tems. Nucleic Acids Research, 34(suppl 1):D689.

Merks, R. M. H., Guravage, M., Inz

e, D., and Beemster, G.

T. S. (2011). Virtualleaf: an open source framework

for cell-based modeling of plant tissue growth and de-

velopment. Plant Physiology, 155(656):666.

Pastore, S. (2006). Web content management systems: us-

ing plone open source software to build a website for

research institute needs. In Digital Telecommunica-

tions, , 2006. ICDT ’06. International Conference on,

page 24.

Sivakumaran, S., Hariharaputran, S., Mishra, J., and Bhalla,

U. (2003). The Database of Quantitative Cellular

Signaling: management and analysis of chemical ki-

netic models of signaling networks. Bioinformatics,

19(3):408.

Weibel, S., Kunze, J., Lagoze, C., and Wolf, M. (1998).

Dublin core metadata for resource discovery. Internet

Engineering Task Force RFC, 2413.

Yu, T., Lawson, J. R., and Britten, R. D. (2009). A dis-

tributed revision control system for collaborative de-

velopment of quantitative biological models. In Mag-

jarevic, R., Lim, C. T., and Goh, J. C. H., editors, 13th

International Conference on Biomedical Engineering,

volume 23 of IFMBE Proceedings, pages 1908–1911.

Springer Berlin Heidelberg.

Yu, T., Lloyd, C., Nickerson, D., Cooling, M., Miller, A.,

Garny, A., Terkildsen, J., Lawson, J., Britten, R.,

Hunter, P., et al. (2011). The Physiome Model Repos-

itory 2. Bioinformatics.

A WEB-BASED REPOSITORY OF REPRODUCIBLE SIMULATION EXPERIMENTS FOR SYSTEMS BIOLOGY

141