Speaker
Mr
Yassene Mohammed
(MediGRID)
Description
MediGRID - Medical Grid Computing
Sax U1, Viezens F1, Mohammed Y1, Lingner Th2, Morgenstern
B2, Vossberg M3,
Krefting, D3, Rienhoff O1
1Department of Medical Informatics, University of Göttingen,
Germany
2Department of Bioinformatics, University of Göttingen, Germany
3Department of Medical Informatics, Charité, Humboldt
University Berlin,
Germany
fred.viezens@med.uni-goettingen.de
Introduction
The project MediGRID [1] combines research institutes from
various areas of
Medicine, Biomedical Informatics, and other Life Sciences.
Numerous associated
partners from industry, healthcare and scientific
institutions ensure a broad
representation of this large community. The main goal of
MediGRID is the
development of a grid-middleware-platform with Globus
Toolkit 4 as a basis for
eScience Services for the community and to help researchers
to use these
services. Here we present the first results of the projects
consisting of the
infrastructure, first community applications and basic
privacy rules.
Materials and Methods
Concerning the data flow in grids, most projects are similar
in the lower layers
(s. Fig. 1), but the biomedical community has to face
particular challenges in the
upper layers. The top layer represents the heterogeneous
biomedical data
sources. Beyond the problem to find the relevant data sets
via metadata
description, access control to the data is of paramount
importance, as the
owner of the data are foremost the patients. Due to the
heterogeneity of the
data we need an additional ontology layer to homogenize the
data. Given
semantic interoperability the researcher can correlate and
analyze the data
with biomedical informatics methods. Biomedical data in
medical grids are
heterogeneous, contain different kinds of information and
have different levels
of privacy. The data might include information about [2]:
• Population: Epidemiology
• Diseases: Clinical practice, clinical trials
• Patient data: Health record, clinical history, physical
exams,
lab/imaging studies
• Organ/tissue: pathology
• Cellular: histology
• Molecular: genetic test results and genomic data.
Having these data online with the suitable tools to connect,
combine and
analyze creates new challenges for data protection and privacy.
Results
The current privacy concepts do not cover the aspects and
abilities of grid
computing. Especially the re-identification risk with the
combination of different
data types has to be assessed. There are severe privacy
concerns related to
genomic-wide association studies [3-5]. These are the main
reasons for the
development of an enhanced security concept for MediGRID. Four
methodological modules are responsible to construct the
suitable
infrastructure: ontology, resource fusion, middleware and
eScience. On the
other hand, three research modules take the initiative to
use this national grid
infrastructure to assist their work: biomedical Informatics,
image processing
and clinical research. The MediGRID consortium developed a
middleware
component as a connector for the medical community to
resources of the
integration project in D-Grid [6].
The community modules use this middleware to “gridify” their
applications in
order to show the advantage of grid computing in medicine.
Ontology
Using OGSA-DAI as a standard of Data Access and Integration
in grids, the
ontology module has successfully developed an ontology tool
and implemented
it as a first step to be a Gridsphere-Portlet in the
MediGRID portal being
available for all project partners.
Bioinformatics
Dialign is a widely used software tool for multiple
alignments of nucleic acid and
protein sequences. Within the MediGRID portal a parallelized
version of the
software is used to speed up the computationally expensive
procedure. In that
way distributed computing allows the user to obtain
high-quality alignments of
bigger databases and longer sequences. AUGUSTUS is a program
that predicts
gene structures in eukaryotic genomic sequences with high
accuracy. Since
AUGUSTUS performs a successive analysis of overlapping
sequence sections, it
is easy to parallelize. Therefore users benefit from
distributed computing with
several instances of the program. Grid resources also allow
frequent update
and centralized storage of huge EST databases.
Image Processing
A 3D image reconstruction for prostate biopsy will register
the different
ultrasonic scans helping the physician to have a new viewing
of the prostate
that was not possible using the traditional 2D methods. The
virtual vascular
surgery helps to calculate and present the animated 3D blood
flow field in the
brain vessels, which could be used to anticipate the
pressure on the walls of
the vessels and for example to predict a bleeding risk. The
Brain 4D MRI image
processing application in MediGRID assists the
identification of the brain areas
research. All three applications are demanding the
possibility of massive data
volume storage and processing. Because of the constant
increase of the data
volume coherent with the development in the imaging
techniques, a dynamic
extendable computing and storage infrastructure is needed,
which ideally will
be a grid environment.
Privacy
In the first step we are working with non person-related
data and set up the
first generation of privacy rules. As a preparation for the
next phase – dealing
with person related data – we defined the additionally
necessary advanced
security methods like audit and tracking. Furthermore some
legal issues
concerning virtual organizations and the ownership of data
and material have
to be addressed.
Further perspective
As we gained experience with our first medical grid
applications, we will be able
to “gridify” other community applications more easily. The
initial incarnation of
the infrastructure is set up; some details are still to
come. As genotyping
constantly gets cheaper, many formerly phenome-related
projects around
complex diseases consider genotyping within the next couple
of years. Beyond
the indisputable opportunities of these studies there are
quite some challenges
to be faced. MediGRID addresses issues like the
homogenization of
heterogeneous data sources and how do we deal with the
well-known privacy
problems. Solving those issues will enhance the portability
of life science grid
services to other projects and other communities.
References
1 www.medigrid.de
2 Martin-Sanchez, F., V. Maojo, and G. Lopez-Campos,
Integrating genomics
into health information systems. Methods Inf Med, 2002.
41(1): p. 25-30.
3 Butte, A.J. and I.S. Kohane, Creation and implications of
a phenome-genome
network. Nat Biotechnol, 2006. 24(1): p. 55-62.
4 Lin, Z., A.B. Owen, and R.B. Altman, Genetics. Genomic
research and human
subject privacy. Science, 2004. 305(5681): p. 183.
5 Kohane, I.S. and Altman R.B., Health-Information Altruists
— A Potentially
Critical Resource. NEJM, 2005. 353 (19): p. 2074-2077
6 www.d-grid.de
This work was supported by the D-Grid Project MediGRID,
funded by the
Federal Ministry of Education and Research (BMBF), FKZ
01AK803A-H.
Author
Prof.
Ulrich Sax
(MediGRID)
Co-authors
Mr
Fred Viezens
(MediGRID)
Mr
Yassene Mohammed
(MediGRID)
