Today, biotechnology needs new methods to make use of the existing scientific results. The electronical infrastructure of the Internet enables us to access remote databases, where the related data is localized. At the same time we can show others the result of our work in a representative way. Referring to molecular biology, there are lots of specific databases available by the WWW (storage layer), followed by already implemented methods to query and analyse the contained data (application layer). The situation is, that because of upcoming semantic, structural and organizational heterogenities generally not all of the recommended data is stored in one common database. To cover the very complex procedures of metabolic networks as we try to do it in our MARGBench project, it's finally necessary to access more than only one database. Our opinion is, that the use of fundamentals of the federated database approach [Conrad, 1997] combined with modern internet programming technology is suitable to solve this task.
Our prototype called MARGBench has been developed to logically integrate distributed molecular data from different databases. The application MetabSIM, a discrete simulation environment, is a reference application to show the functionality of MARGBench. In the following, the architecture of the prototype will be introduced. Afterwards the simulation model of MetabSIM will be presented.
MARGBench has been throughout implemented in the Java programming language, which is why the components of the system can be installed on any platform where the Java Runtime Environment is available. Until now, we have been successfully tested on Microsoft Windows and Sun Solaris. Figure 1 displays the global architecture of MARGBench and the connection of MetabSIM to the systems interface. Instead of MetabSIM any other application can be docked. The MARGBench-architecture consists of the three main components:
The BioDataServer is the core of the system and is connected to the component databases (BRENDA [Schomburg, 1998], TRANSFAC [Wingender et al., 1997], ...) by specialized database adapters [Stephanik, 1999]. The seperate adapters can be manufactured automatically by an adapter-generator for flat/semi-structured files, multiple types of database systems and Internet information systems. After processing the adapters, they can be bound to the BioDataServer at runtime. The BioDataServer manages user-defined integrated schemes. Here the user describes, which spectrum of data should be extracted from every adapter. The BioDataServer accepts SQL-statements to query the data.
|
Figure 1: Software-architecture of MARGBench. |
The main advantage of the BioDataServer is it's flexibility. Every user can specify his own scheme. The client/server-communication relies on socket-communication. But the speed of processing the queries is depending from the quality of the Internet connection between BioDataServer and component database as well as from the integration costs. This can be to slow for huge amounts of data, which is why a proxy-cache is needed. As next, the query-results are in string format and must be transformed into the applications data structure.
The MARGBench Factory hides the BioDataServer from the user and offers a CORBA based interface dealing the main aspects of metabolic networks. Once data from the BioDataServer is read, it will be stored in the underlying object-oriented database in the MARGBench-Factory. Proceeding from a BioDataServers user-defined scheme an IDL-description is produced. Using this thin-client-interface any client application can access objects from the MARGBench-Factory, and work with them as references in their own scope. The real data is positioned in the server at any time. As it is the characteristic of CORBA, the object references can be exchanged between CORBA-applications via Internet.
Developing DBMS-supported applications forces the programming of database-related component to establish database-connection, query the data, transmit the results, store the data and so on. The MARGBench-Browser provides this functionality and can be included as component in different Java-applications. To verify the main idea of MARGBench, our workgroup is developing an own simulation application, MetabSIM.
The simulation environment MetabSIM has been designed to animate metabolic networks and counts to the group of discrete simulations. The formalism of the MetabSIM-grammar bases on the approaches of [Hofestädt, 1996] and [Meineke, 1994]. The choose of a discrete method is arising from the generally high complexity of biological regulatoric procedures.
 |
Figure 2: Example of biocatalytical reactions. |
In MetabSIM the discrete states of cellular compartments are expressed by mixtures of substrates. The states are called configurations. The step from one configuration to a following (derivation) is defined by a rule-set, where each rule describes a biochemical process abstractly.
A metabolic rule is an 8-Tuple r = (b, g, c, f, V, N, F, H) with the elements:
b basic probability,
g basic speed,
c basic speed constant,
f basic behavior,
V substrates,
N products,
F enhancing influences and
H inhibiting influences.
Figure 2 displays the application of a hypothetical rule Substrate ® Product under the positive influence of an Enzyme, which is regulated by an effector. Because of the symbolic character of the example only a poor resolution is choosen. The enhancing effect is visualized once more in Figure 3.
 |
Figure 3: The influence of an allosteric effector. |
Rules are not only to describe simple chemical reactions. They allow to model chains of reactions, where data is only marginally available, as "Black Box". Finally, the regulatoric effect benefits from cascading influence functions. The data to build up rules and the values to describe the substance-behavoir is directly taken from the MARGBench-system.
MARGBench is an information system specialized to be applied in the area of molecular biotechnology. The system provides services to integrate the data of different databases and to cache a part of the data locally. Two interfaces, relational-uncached and object-oriented-cached are available. The interfaces use standardized query languages (SQL/OQL) to query the data.
As a reference software, the tool MetabSIM demonstrates the capabilities of MARGBench in view of metabolic networks. MetabSIM is able to animate regulatoric effects in discrete states of cellular compartments.
Some components of our system rely on former prototypes and are still objects of our current work. Our roadmap points to extend the number of accessed databases, to automate the adaptation to special cases of application and to improve the simulation model step by step.