A METHODOLOGY FOR INTEGRATING NEW SCIENTIFIC

DOMAINS AND APPLICATIONS IN A VIRTUAL LABORATORY

ENVIRONMENT

E. C. Kaletas, H. Afsarmanesh, L. O. Hertzberger

University of Amsterdam, Informatics Institute

Kruislaan 403, 1098 SJ Amsterdam, The Netherlands

Keywords:

Virtual laboratory, scientiﬁc domain integration, scientiﬁc application integration, integration methodology

Abstract:

Emergence of advanced complex experiments in experimental sciences resulted in a change in the way of

experimentation. Several solutions have been proposed to support scientists with their complex experimen-

tations, ranging from simple data portals to virtual laboratories. These solutions offer a variety of facilities

to scientists, such as management of experiments and experiment-related information, and management of

resources. However, issues related to adding new types of experiments to proposed support environments still

remain untouched, causing inefﬁcient utilization of efforts and inadequate transfer of expertise. The main topic

of this paper is to present a methodology for integrating new scientiﬁc domains and applications in a multi-

disciplinary virtual laboratory environment. In order to complement the methodology with the right context,

the paper also presents an experiment model that uniformly represents scientiﬁc experiments, data models for

modelling experiment-related information, and mechanisms for the management of this information.

1 INTRODUCTION

In recent years, technological advances and system-

atic growth of research efforts in experimental science

domains, such as life sciences, have achieved break-

throughs that go beyond the imagination of mankind

even a decade ago. Advances in the laboratory tech-

niques and the introduction of highly automated lab-

oratory instruments have caused both the automation

of many steps in scientiﬁc experiments and the gen-

eration of very large amounts of data, leading to a

change in the way that scientists perform their sci-

entiﬁc research. However, new ways of experimenta-

tion pose many requirements and challenges to both

scientists and developers of supporting tools and in-

frastructures. The following can be mentioned among

these challenges, which create a burden on the shoul-

ders of scientists and developers of supporting tools:

• Increasing number and complexity of laboratory

instruments and procedures require signiﬁcant ef-

fort to perform an experiment. For example, clone

preparation phase of a microarray experiment con-

tains up to 51 steps.

• Large amounts of data are generated during experi-

ments. For instance, material analysis experiments

for complex surfaces typically generate 20 images

per day where data size goes up to 100 MB per im-

age (Frenkel et al., 2001).

• Heterogeneity causes difﬁculties in modelling and

storage of experimental information. For instance,

281 biological database resources in 18 different

categories were listed in 2001 (Baxevanis, 2001).

• Lack of standards for modelling and representation

of information causes wasted efforts for developing

speciﬁc solutions to overcome common problems

and difﬁculties in comparing the results obtained

in multiple experiments.

New solution methods and problem solving tech-

niques are needed to address the challenges. How-

ever, addressing these challenges require long-term

multi-disciplinary efforts between scientists from

both experimental science domains and computer sci-

ence. In experimental science domains, scientists in-

volved in these experimentations must both advance

their models and mechanisms supporting the analy-

sis of resulted information, and design innovative in-

vestigation approaches leading to discoveries. From

the computer science point of view, however, the

complexity of emerging experimentations necessi-

tates that the support environment be generic and con-

stitute a horizontal base infrastructure, on top which

265

C. Kaletas E., Afsarmanesh H. and O. Hertzberger L. (2004).

A METHODOLOGY FOR INTEGRATING NEW SCIENTIFIC DOMAINS AND APPLICATIONS IN A VIRTUAL LABORATORY ENVIRONMENT.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 265-272

DOI: 10.5220/0002634802650272

 SciTePress

vertical software tools can be developed, each provid-

ing a speciﬁc service for the experimentations.

Several solutions have been proposed to sup-

port scientiﬁc experimentations, ranging from simple

Web-based query interfaces for scientiﬁc databases

to complex problem solving environments and vir-

tual laboratories. Some of these environments target

speciﬁc applications while others are offering more

generic support. However, issues such as extension of

support environments with facilities for emerging sci-

entiﬁc experiments (i.e. integration of vertical soft-

ware tools in the horizontal support environments)

still remain untouched. More precisely, a method-

ology for integrating new scientiﬁc domains and ap-

plications in a support environment is still lacking;

causing inefﬁcient utilization of efforts and inade-

quate transfer of expertise. Yet, a further challenge

would be to provide a methodology for integrating di-

verse heterogeneous scientiﬁc applications in a multi-

disciplinary support environment.

The main topic of this paper is to present a method-

ology for integrating new scientiﬁc domains and ap-

plications in a multi-disciplinary virtual laboratory

environment. Nevertheless, integration of new do-

mains and applications in a complex environment like

virtual laboratory requires a common understanding

of scientiﬁc experiments, availability of both well-

deﬁned data models for modelling experiment-related

information and mechanisms for the management of

this information. The remaining of this paper is struc-

tured as follows: First an overview of existing and

emerging support environments is provided in Sec-

tion 2. In Section 3, general structure of scientiﬁc

experiments is presented, followed by a description

of an experiment model. Then ‘process data ﬂows’

are introduced in this section as a tool for modelling

experiments. Section 4 focuses on a speciﬁc support

environment, namely the VLAM-G virtual laboratory

environment. Section 5 further focuses on VIMCO,

the information management platform of VLAM-G,

and addresses issues related to modelling and man-

agement of information related to scientiﬁc experi-

ments. The proposed methodology is described in

Section 6. Finally, Section 7 concludes.

2 SUPPORT ENVIRONMENTS

This section describes different types of support envi-

ronments for scientiﬁc experimentations. The remain-

ing of this paper, however, focuses on virtual labora-

tories.

Science Portals. Science Portals (Ashby et al.,

2001), (Pierce et al., 2002) are emerging as conve-

nient mechanisms for providing a single point of ac-

cess with familiar and simpliﬁed interfaces to a spe-

ciﬁc set of resources that are of importance to a spe-

ciﬁc scientiﬁc community. Resources can be, for in-

stance, computational, storage, networking resources,

electronic whiteboards, or a digital library. These re-

sources are usually made available to users as ser-

vices. Users can make use of these services either

in a stand-alone manner using only one service at a

time, or in a collective manner using a number of ser-

vices simultaneously. Science portals only provide a

uniform means for accessing resources. Users them-

selves are responsible for the correct and efﬁcient us-

age of the available resources.

Problem Solving Environments. A Problem Solv-

ing Environment (PSE) (Allen et al., 2001),

(Schuchardt et al., 2002) is a computer system that

provides all the computational facilities needed to

solve a target class of problems (Gallopoulos et al.,

1994). Common characteristics of PSEs include

among others a target class of science or engineer-

ing problems, natural appearance and ease of use,

provision of multiple solution paths or algorithms in

the target areas, and availability of parameterized al-

gorithms. Users can be assisted through automatic

and/or semiautomatic selection of a proper solution

method from the available set of solution methods, or

by providing ways to easily incorporate new solution

methods. Although some examples of generic PSEs

do exist such as Cactus (Allen et al., 2001), most of

the current PSEs target a speciﬁc class of problems.

Virtual Laboratories. A Virtual Laboratory (VL)

(Afsarmanesh et al., 2002), (Messina, 2003) provides

an electronic workspace for distributed collaboration

and experimentation in research, to generate and de-

liver results using distributed information and com-

munication technologies (Vary, 2000). It supports

an aggregation of people who pursue a related set

of research activities and share resources, where the

resources including the people may be geographi-

cally distributed and associated with different institu-

tions. Therefore, virtual laboratories bring together

best combination of skills, expertise, and tools to

carry out the same type of research that is done in

a single real laboratory (Messina, 2003).

At present, the concept of ‘virtual laboratory’ has

been associated with several different meanings, de-

pending on the wide variety of application interests

and ﬁelds (e.g. education, games, science, engineer-

ing, manufacturing, etc.). On the other hand, there are

some recent VL projects that target the provision of

generic solutions and offer a wide variety of function-

ality, and that are applicable to a wider set of problems

(Afsarmanesh et al., 2002), (Catlin et al., 2000). One

such virtual laboratory that offers generic solutions to

ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

266

scientists is the Grid-based Virtual Laboratory Ams-

terdam (VLAM-G), which is described in Section 4.

3 MODELLING EXPERIMENTS

In a VL environment, it is natural that all activities are

centered around experiments. The experiment model

underlying a VL outlines the overall approach that

the VL follows for supporting scientiﬁc experiments.

Furthermore, the experiment model allows a method-

ological deﬁnition of complex experiments in a prob-

lem domain. Thus, the experiment model lies in the

core of a VL. This section focuses on the modelling

of scientiﬁc experiments and describes an experiment

model, by ﬁrst presenting high-level general structure

of scientiﬁc experiments, then providing a description

of the experiment model. This section then introduces

‘process data ﬂows’ as a tool for modelling experi-

ments.

3.1 General Structure of

Experiments

A scientiﬁc experiment consists of a number of ex-

periment components. Components of an experiment

correspond to activities, physical entities, or data ele-

ments involved in that experiment. Every experiment

is part of a project, which may contain one or more

experiments that are related to each other (Figure 1).

Project

Experiment

Experiment Component

Figure 1: General structure of scientiﬁc experiments

A project is a grouping of related experiments. Ex-

periments in a project may follow an order (e.g. time

course experiments), or may be unordered (e.g. com-

parison experiments).

Experiments are composed of three kinds of exper-

iment components (see Figure 2 for experiment com-

ponents involved in a typical microarray experiment).

Physical entities are used during laboratory activities

or during instrumentation. Activities may represent

laboratory activities, instrumentations, or computa-

tional processes. Data elements in turn correspond

to both raw data generated by instruments and input

data to computational processes. The generic descrip-

tive elements represent the components that are com-

mon to all experiments and provide descriptive infor-

mation about them, such as organism, gene, or infor-

mation about hardware and software tools (e.g. scan-

ner). Such descriptive information does not change

from one experiment to another.

HYBRIDIZATION CLONE

HYBRIDIZED

ARRAY

ARRAY IMAGEMICROARRAY

ARRAY

MEASUREMENT

ARRAY

SCANNING

ARRAY IMAGE

ANALYSIS

GENE

ORGANISM

IMAGE

ANALYSIS

PROGRAM

SCANNER

NAME

Notation:

Physical Entity

Activity

Data Element

Generic Descriptive

Element

SAMPLE

MRNA PROBE

Figure 2: Components for microarray experiments

3.2 The Experiment Model

An experiment model is developed to uniformly rep-

resent scientiﬁc experiments from different domains.

A brief summary of the experiment model is pro-

vided below. For more information on the experiment

model, please refer to (Kaletas et al., 2003).

The experiment model consists of two main com-

ponents, namely experiment procedure and experi-

ment context (Figure 3).

Experiment Procedure

Experiment Context

Experiment

Figure 3: The experiment model

An experiment procedure deﬁnes the approach

taken to solve a particular scientiﬁc problem, by

deﬁning the experiment components that are typically

involved in the experiments of the same type. Thus,

an experiment procedure standardizes the experimen-

tal approach for experiments of the same type.

An experiment context, on the other hand, de-

scribes the solution. An experiment context is an in-

stantiation of an experiment procedure. It describes

the accomplishment of a particular experiment, by

providing descriptions of each experiment component

involved in the experiment. Thereby, it provides the

context for that particular experiment.

Processing and analysis of large data sets consti-

tute an important part of scientiﬁc experiments, and

hence require special attention. In an experiment, dur-

ing its processing and analysis, data ﬂows from one

computational process to another. This data ﬂow is

represented by a directed graph. Nodes in the graph

are computational processes, while the connecting

arcs are the data ﬂowing through the processes. This

A METHODOLOGY FOR INTEGRATING NEW SCIENTIFIC DOMAINS AND APPLICATIONS IN A VIRTUAL

LABORATORY ENVIRONMENT

267

data ﬂow graph is called as computational processing.

An experiment procedure may contain computational

processes. Consequently, particular experiments that

are instantiating this procedure contain descriptions

of these processes in their experiment contexts.

3.3 Process Data Flows

Given a scientiﬁc domain, the ﬁrst step in modelling

the experiment-related information is to perform a de-

tailed analysis of the experiment components for each

experiment in that domain. Result of the analysis is a

conceptual model, which includes a comprehensive

component-by-component deﬁnition of experiments

in the domain.

In this conceptual model, every component in-

volved in an experiment, the properties of each com-

ponent, and the relationships among the components

are deﬁned. Such a detailed and comprehensive def-

inition of an experiment is referred to as the Process

Data Flow (PDF). Figure 4 presents the microarray

PDF as an example using the same notation as Fig-

ure 2 for experiment components. This example mi-

croarray PDF depicts the order of components in the

experiment (i.e. the experiment ﬂow) and also the re-

lationships between components and generic descrip-

tive elements in microarray experiments.

HYBRIDIZATION

CLONE

HYBRIDIZED

ARRAY

ARRAY IMAGE

MICROARRAY

ARRAY

MEASUREMENT

ARRAY

SCANNING

ARRAY IMAGE

ANALYSIS

GENE ORGANISM

IMAGE

ANALYSIS

PROGRAM

SCANNER

NAME

Notation:

ARRAY

SPOTTING

ARRAY

SPOTTER

ARRAY

SPOTTER

PROGRAM

MRNA PROBE

MRNA

LABELING

MRNA

EXTRACTION

TREATED

CELL SAMPLE

TREATMENT

CELL SAMPLE

SLIDE

PROCESSOR

SPOT

Physical Entity

Activity

Data Element

Generic Descriptive

Element

Experiment flow /

Relationship between

experiment components

Relationship between an experiment

component and a generic descriptive element

Relationship between two

generic descriptive elements

Figure 4: Process-Data Flow for microarray experiments

4 VLAM-G

The Dutch VLAM-G (Grid-based Virtual Laboratory

Amsterdam) project (Afsarmanesh et al., 2002), (Af-

sarmanesh et al., 2001) provides the main context

for the work presented in this paper. VLAM-G is

a multi-disciplinary virtual laboratory environment

that provides the required generic environment for

multi-disciplinary research in experimental science

domains.

The two initial application cases of VLAM-G were

the DNA microarray application from the life sci-

ences domain and the material analysis for complex

surfaces application (MACS) (Frenkel et al., 2001)

from the physics domain. The work described in this

paper has been initiated and motivated by the need

to support the requirements of both these initial ap-

plication cases and any future applications within the

VLAM-G.

VLAM-G allows its users to:

• perform multi-disciplinary, collaborative experi-

ments in a uniform, integrated environment,

• complement their in-vitro experiments with in-

silico experiments,

• deﬁne customized experimental procedures and

analysis ﬂows,

• reuse generic software components, and

• share hardware, software, storage, networking re-

sources as well as knowledge and experience.

The architecture of the VLAM-G and interaction

among its components are shown in Figure 5. Main

components of the architecture are described below.

user

Front-End

Collaboration

Session Manager

RTS

Assistant

VIMCO

Grid / Globus services

M1 M2 M3 M4

Module Repository

Resource A Resource B

…

RTS

Project

VIMCO

…

Application

DBs

Figure 5: VLAM-G architecture

Front-End is the user environment of the VLAM-

G. Users interact with the VLAM-G only through the

Front-End. The Front-End consists of a number of

graphical user interfaces (GUIs), which present the

VLAM-G functionality to its users in a uniform way.

Session Manager acts as a gateway to the VLAM-

G server side. It manages the active user sessions, and

is responsible for coordinating the interactions among

the VLAM-G components.

The distributed computing and networking re-

sources on the Grid are made available to VLAM-

G users through the Run Time System. Generic

ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

268

VLAM-G processing functionality and application-

speciﬁc problem solvers are presented to its users as

self-contained software entities. RTS provides an API

which encapsulates the Grid computing code within a

simple interface. Users compose their computational

processings by attaching a number of software entities

to each other, which are then executed by the RTS on

the Grid environment.

Collaboration component of the VLAM-G en-

ables simultaneous collaborative design and execu-

tion of experiments. The Assistant assists users dur-

ing the design of an experiment, for instance, by sug-

gesting the most efﬁcient software to perform a spe-

ciﬁc task. Implementation of these two components

is planned for future work.

Module Repository is a persistent storage for bi-

naries of software entities.

The Virtual Laboratory Information Management

for Cooperation (VIMCO) is the information man-

agement platform of VLAM-G. VIMCO is described

in details in the following section.

5 VIMCO

This section describes the VIMCO facilities for mod-

elling and manipulation of information related to sci-

entiﬁc experiments in the VLAM-G environment.

5.1 Modelling Information Related

to Scientiﬁc Experiments

Modelling experiment-related information handled in

a VL is important, since the employed data models

have a direct inﬂuence on the architectural compo-

nents of the VL and their functionalities. Information

related to scientiﬁc experiments include experiment

information (e.g. experiment procedures, experiment

contexts, computational processings), user informa-

tion (e.g. information about users and user roles, ac-

cess rights deﬁnitions), and VL-related information

(e.g. session information, information about the ser-

vices provided by the VL). This section builds on the

experiment model presented earlier in this paper, and

describes the modelling of information related to sci-

entiﬁc experiments in VL.

5.1.1 Experiment Information

In this subsection, data models developed for the

modelling of different types of information about sci-

entiﬁc experiments are described. Note that all data

models described below are object-oriented. Due to

space limitations, however, UML diagrams and de-

tailed descriptions of the data models are omitted. De-

tails of the data models presented here can be found

in (Kaletas et al., 2003).

Modelling Experiment Procedures. The

Procedure Data Model facilitates the rep-

resentation and storage of experiment pro-

cedures, and consists of the following

classes: Procedure, ProcedureElement,

ProcedureConnection, and ProcedureGUI.

An instance of the Procedure class contains the

general information about a procedure, including its

version. Multiple versions of the same procedure can

exist. Every Procedure object consists of a set of

ProcedureElements, which correspond to exper-

iment components. For each procedure element (and

hence experiment component), there is a correspond-

ing data type in the application database schema to

hold the information about that experiment compo-

nent. The corresponding data type in the database

schema for a procedure element is represented by the

className attribute. isProcessing attribute of

a procedure element indicates whether the element

actually corresponds to a processing. Instances of

ProcedureConnection class correspond to

the relationships between classes in the database

schema, indicated by the relName attribute. Finally,

ProcedureGUI objects keep information about the

graphical properties of a procedure for displaying

it to users. Procedure deﬁnitions are stored in the

application databases.

Modelling Experiment Contexts. A context is an

instance of a procedure, and the procedure elements

correspond to data types in an application database.

This means, a context actually consists of a number of

objects in the application database, hence, modelling

the application database schema for the domain of an

experiment is sufﬁcient for modelling the contexts of

that experiment.

Using the general structure of scientiﬁc experi-

ments, a base data model for experimental infor-

mation called Experimentation Environment Data

Model (EEDM) is designed. EEDM covers the com-

mon aspects of experiments; namely scientist owning

the experiment, input, activities applied on the input,

output obtained from an activity, hardware and soft-

ware used during the experiment, conditions and pa-

rameters for the activities, devices and software, and a

recursive ﬂow of processes and data where a speciﬁc

order is followed during an experiment.

EEDM provides generic constructs for these com-

mon aspects of scientiﬁc experiments, which are ex-

tended (sub-typed) in application databases for mod-

elling experiment-speciﬁc information.

A METHODOLOGY FOR INTEGRATING NEW SCIENTIFIC DOMAINS AND APPLICATIONS IN A VIRTUAL

LABORATORY ENVIRONMENT

269

Modelling Software Entities and Computational

Processings. A computational processing covers

both descriptions of the executables used (i.e. soft-

ware entities) and run-time information for the pro-

cessing. Software entities represent self-contained ex-

ecutable programs. They are mainly characterized by

the tasks that they perform, their input and output data

types, parameters, and run-time requirements. For ev-

ery executable, its developer speciﬁes the run-time re-

quirements to help a user to decide which software

entities to use in his/her processing. A user designs

a computational processing by attaching a number of

software entities to each other through their input and

output ports. This design is stored in RTS DB for fu-

ture references and for possible re-runs.

5.1.2 User Information

User information is used in a VL for security and

access control, and for administration purposes. In

addition, application databases contain user informa-

tion to capture, for instance, the owner of a physical

entity or data element (e.g. microarray or array im-

age), or operator of an instrument (e.g. array scan-

ner). As such, information about VL users is deﬁned

and stored in all VL databases.

Every VL user has a unique username, and assumes

a role. Roles are used to deﬁne and enforce access

rights for groups of users. Access to information in

the VL is controlled through a set of restrictions. It

is considered that by default every user has all privi-

leges on all available information. Depending on the

role that a user assumes, these unlimited privileges are

restricted by a set of ‘restrictions’. Below, the speciﬁc

roles deﬁned for VL users are summarized.

• Application users are the actual users of the VL.

An application user is typically associated with a

scientiﬁc domain (e.g. molecular biology), and can

only access the application database for this do-

main. S/he has full control on her/his contexts and

processings but only read access to procedures.

• Domain experts are application users who have

extensive knowledge and experience in a given do-

main and on the experiments being performed in

that domain. They are responsible for modelling

application databases (together with Information

Management Admin), designing experiment proce-

dures, deﬁning laboratory protocols, and register-

ing descriptions of software entities.

• Tool developers are the developers of software en-

tities (i.e. vertical software tools).

• Administrators are responsible for the proper

management and operation of the VL. Informa-

tion Management Admin is responsible for mod-

elling application database schemas and for main-

taining these databases. VL Admin is responsi-

ble for resource management, infrastructure main-

tenance, and user management.

5.1.3 VL-Related Information

During its operation, the VL itself manages and ma-

nipulates certain types of information, such as infor-

mation about active sessions and information about

the services it provides.

Session management is an important functionality

in a VL for controlling and coordinating user activi-

ties. All user activities occur within a session. For re-

covery reasons, VIMCO provides persistence to ses-

sion information (in VIMCO DB). A Session class

is deﬁned to store the session information, which con-

tains owner of the session, observers in case of a col-

laborative session, and the procedures, contexts, and

processings being used in the session.

The services that are made available by the VL

to its users are highly dependent on the existing

experiment-related information in the VL. For in-

stance, availability of a ‘microarray experiment pro-

cedure’ allows a VL to provide ’making microarray

experiments’ as a service to its users. In order to pro-

vide a list of services to users, VL makes use of the

experiment-related information stored and managed

by VIMCO. The list of available services is generated

by VIMCO by extracting the necessary information

from what is available in the databases.

5.2 Managing Information Related

to Scientiﬁc Experiments

VIMCO provides the necessary functionality and

mechanisms for the manipulation of different types

of information that it manages. Very brieﬂy, VIMCO

supports the following functionality:

• managing experiment procedures, contexts, and

computational processings;

• deﬁning and enforcing usage policies for experi-

ment contexts;

• managing user information;

• deﬁning and enforcing access rights;

• managing VL-related information;

• uniformly accessing multiple, heterogeneous appli-

cation databases; and

• sharing and exchanging information among collab-

orating partners.

ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

270

6 THE PROPOSED

METHODOLOGY

Lack of a methodology for supporting new scien-

tiﬁc domains and applications in a virtual laboratory

causes inadequate transfer of expertise and hence in-

efﬁcient utilization of efforts. In this paper, sev-

eral different aspects of a virtual laboratory environ-

ment were addressed with a focus on its informa-

tion management component. In speciﬁc, an exper-

iment model is presented for uniformly representing

diverse scientiﬁc experiments. Furthermore, several

data models for modelling different types of informa-

tion related to scientiﬁc experiments, and function-

ality for the actual management of this information

are provided. The results obtained and the experience

gained during the VLAM-G project, and especially

during the two application cases, are used for deﬁning

a methodology for integrating new scientiﬁc domains

and applications in VL. The methodology provides

step-by-step guidance to domain experts, tool devel-

opers and administrators during the process of adding

new domains, applications and resources to VL.

The methodology provided in this section is

generic and reusable; it can be uniformly applied to

different scientiﬁc domains and to different scientiﬁc

applications. This is partially due to the genericness

and reusability of the models described here.

Following subsections describe the methodology.

6.1 Integration of a New Domain

Integration of a new domain in VL covers the devel-

opment of a domain-speciﬁc application database, in-

tegrating and registering this database into the VL and

its information management platform, and deﬁning

and registering the domain users (application users).

Steps to take when integrating a new domain in VL

are enumerated below.

1. Study different types of experiments being per-

formed in the domain, to the level of detail of single

components involved in the experiments and the re-

lated information for each component. This step is

realized by the experts in the domain, in close col-

laboration with an Information Management Ad-

min (IM Admin).

2. Generate process data ﬂows (PDFs) for each ex-

periment type. Discuss the PDFs with the experts

from the domain, make any necessary modiﬁca-

tions, and eventually conﬁrm the PDFs. This step

is performed by a Domain Expert in close collabo-

ration with an IM Admin.

3. For each PDF, map the PDF elements to the base

data types in the EEDM, and model each PDF el-

ement by sub-typing one of the base EEDM data

types. The result is the process data ﬂows with ele-

ments that are deﬁned as sub-types of base EEDM

types. These PDFs form the base for the applica-

tion database schema for the domain. This step is

performed by an IM Admin together with a Do-

main Expert.

4. Cross-study different experiment types deﬁned in

terms of extended EEDM constructs, and identify

the elements that are common to different exper-

iments (e.g. organism info). Identify the relation-

ships between components in different experiments

and these common elements. Conﬁrm the result

with experts in the domain. The result is the core

schema for the application database. This step is

performed by a Domain Expert and an IM Admin.

5. Map the core database schema to a data deﬁnition

language (e.g. Object Deﬁnition Language). Add

any other data models required by the information

management platform (e.g. procedure data model).

This step is performed by an IM Admin.

6. Create the application database by loading the

schema deﬁned in the data deﬁnition language.

This step is performed by an IM Admin.

7. Register the database to the information manage-

ment platform by providing the required informa-

tion about the new data source. This step is per-

formed by an IM Admin.

8. Deﬁne and register the domain users using the user

management functionality. This step is performed

by a VL Admin.

6.2 Integration of a New Application

Once the new domain is integrated, new applications

from that domain can be integrated. Steps to take

when integrating new applications are given below.

1. Deﬁne experiment procedures for different types of

experiments that will be offered to users of this ap-

plication. Graphical editors provided by the VL can

be used for procedure deﬁnition. This activity is

performed by a Domain Expert.

2. Develop software entities for application speciﬁc

functionality (e.g. analysis tool for microarray

data). This step is performed by a Tool Developer.

3. After the quality control applied to software enti-

ties, store their deﬁnitions in the information man-

agement platform, and register them to VL. These

activities are performed by a Domain Expert.

4. Using the access rights management functionali-

ties, deﬁne the user roles and restrictions for each

role on the data types of this domain. Assign one

role to each domain user. These activities are per-

formed by a Domain Expert.

A METHODOLOGY FOR INTEGRATING NEW SCIENTIFIC DOMAINS AND APPLICATIONS IN A VIRTUAL

LABORATORY ENVIRONMENT

271

7 CONCLUSIONS

The main subject of this paper was to provide a

methodology for integrating emerging scientiﬁc do-

mains and applications in a virtual laboratory envi-

ronment. Integration of new domains and applications

in a complex virtual laboratory environment requires

a common understanding of scientiﬁc experiments,

availability of both well-deﬁned data models for mod-

elling experiment-related information and functional-

ity / mechanisms for the management of this informa-

tion. Therefore, in this paper:

1. an experiment model is introduced that is capable

of uniformly representing different aspects of het-

erogeneous scientiﬁc experiments;

2. several data models are provided for representing

the experiment-related information, which are uni-

form and reusable (to model information related to

heterogeneous experiments), and open, ﬂexible and

extendible (to improve the developed models in the

future when needed and to model new types of ex-

perimental information);

3. mechanisms for managing the experiment-related

information in a VL are mentioned, that are generic

and reusable (to support uniform management of

heterogenous experimental information); and

4. a methodology is deﬁned for integration of new do-

mains and applications in VL based on the results

obtained and experience gained during the VLAM-

G project presented here, which provides a step-

by-step guidance to domain experts, tool develop-

ers and administrators during the process of adding

new applications / resources to the VL.

The methodology deﬁned in this paper is generic

and can be implemented by different support envi-

ronments in different ways. During the course of the

VLAM-G project, the methodology has been applied

to the two initial application cases (the DNA microar-

ray and MACS applications), and the received feed-

back was used to further improve/reﬁne the method-

ology. Currently, two databases for these applications

together with experiment procedures and some anal-

ysis tools are provided to VLAM-G users. Further-

more, development of a new application for gene se-

quence analysis studies using this methodology is on-

going.

REFERENCES

Afsarmanesh, H., Belleman, R. G., Belloum, A. S. Z., Ben-

abdelkader, A., van den Brand, J. F. J., Eijkel, G. B.,

Frenkel, A., Garita, C., Groep, D. L., Heeren, R.

M. A., Hendrikse, Z. W., Hertzberger, L. O., Kaan-

dorp, J. A., Kaletas, E. C., Korkhov, V., de Laat, C.

T. A. M., Sloot, P. M. A., Vasunin, D., Visser, A., and

Yakali, H. H. (2002). Vlam-g: A grid-based virtual

laboratory. Scientiﬁc Programming, 10(2):173–181.

Afsarmanesh, H., Kaletas, E. C., Benabdelkader, A., Garita,

C., and Hertzberger, L. O. (2001). A reference archi-

tecture for scientiﬁc virtual laboratories. Future Gen-

eration Computer Systems, 17(8):999–1008.

Allen, G., Benger, W., Dramlitsch, T., Goodale, T., Hege,

H. C., Lanfermann, G., Merzky, A., Radke, T., Seidel,

E., and Shalf, J. (2001). Cactus tools for grid applica-

tions. Cluster Computing, 4(3):179–188.

Ashby, J. V., Bicarregui, J. C., Boyd, D. R. S., van Dam,

K. K., Lambert, S. C., Matthews, B. M., and O’Neill,

K. D. (2001). A multidisicplinary scientiﬁc data por-

tal. In Proceedings of the 9th International Confer-

ence and Exhibition on High-Performance Computing

and Networking (HPCN Europe 2001), pages 13–22.

Baxevanis, A. D. (2001). The molecular biology database

collection: An updated compilation of biological

database resources. Nucleic Acids Research, 29(1):1–

10.

Catlin, A. C., Gaitatzes, M., Houstis, E., Ma, Z., Markus, S.,

Rice, J. R., Wang, N.-H., and Weerwarana, S. (2000).

Enabling Technologies for Computational Science -

Frameworks, Middleware and Environments, chap-

ter 24, pages 301–313. Kluwer Academic Publishers.

Frenkel, A., Afsarmanesh, H., Eijkel, G. B., and

Hertzberger, L. O. (2001). Information management

for material science applications in a virtual labora-

tory. In Proceedings of the 12th International Con-

ference on Database and Expert Systems Applications

(DEXA 2001), pages 165–174.

Gallopoulos, E., Houstis, E., and Rice, J. R. (1994). Com-

puter as thinker/doer: Problem-solving environments

for computational science. Computing in Science and

Engineering, 1(2):11–23.

Kaletas, E. C., Afsarmanesh, H., and Hertzberger, L. O.

(2003). Modelling multi-disciplinary scientiﬁc exper-

iments and information. In Proceedings of the Eigh-

teenth International Symposium on Computer and In-

formation Sciences (ISCIS03), pages 75–82.

Messina, P. (2003). The emergence of virtual laborato-

ries for science and engineering - igrid2002 presen-

tation. Electronically available at: http://www.

igrid2002.org/ppt/Paul_Messina.ppt.

Pierce, M., Youn, C., and Fox, G. C. (2002). The gateway

computational web portal. Concurrency and Com-

putation: Practice and Experience, 14(13–15):1411–

1426.

Schuchardt, K. L., Myers, J. D., and Stephan, E. G. (2002).

A web-based data architecture for problem solving en-

vironments: Application of distributed authoring and

versioning to the extensible computational chemistry

environment. Cluster Computing, 5(3):287–296.

Vary, J. P. (2000). Report of the expert meeting on virtual

laboratories. Technical Report CII-00/WS/01, United

Nations Educational, Scientiﬁc and Cultural Organi-

zation.

ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

272