The University of Iowa Department of Paediatrics co-ordinates an extensive study for autism gene identification. The research fits in with the methods of Genetic Linkage Analysis, one of the applications in the Human Genome Project. In this regard, GenoMap is the very first implementation of a system involving several groups of co-operating users at multiple scientific institutions in a common attempt to isolate the genomic locus of controlling genes in hard to cure diseases, such as autism and cancer. To this purpose, the Parallel Processing Laboratory at Iowa University has developed a whole new class of applications accessible through the World Wide Web in a three tier hierarchy of parallelism, including heterogeneity within a single problem instance, homogeneity among user subgroups within one problem domain, and multiple instances of the entire problem class, sharing the software and hardware computational resources.
In order to determine the association between a genetically-linked trait with reference to autism or cancer, and its human genome locus among approximately three billion base pairs, a tight collaboration between a group of experts like geneticists, clinical physicians, laboratory technicians, disease specialists, statisticians and computer engineers is required. Since about 2 years GenoMap has analysed over 400 genetic loci with data describing traits of more than 300 patients suffering from autism. The system is able to offer a portable, intuitive interface to manage all the data related to the process of gene discovery and location. The core of the architecture consists of a socket server to register clients and servers, and to guarantee the secure isolation of users in separate problem domains.
At the first level, each user needs to view the GenoMap architecture through the "window" of his own part to the collaboration. At the second level, the geographical distance of the users comes into play as well as their growing number. In the autism study, scientists from three universities are involved, namely Iowa, Tufts and Vanderbilt, to collect phenotypes or patient states, and genotypes or the states of the 400 different genome loci, as well as to conduct analyses to define which loci are major candidates for autism. At the third level, the computational resources are distributed allowing the users to efficiently share information and software in such a way that data privacy is being secured. The paradox of conflicting demands posed by heterogeneity and security needs is solved by identity verification.
GenoMap has been implemented in Java, featuring applets which secure the access to the tools by means of passwords. Users are only allowed to retrieve contents from the databases within the context of a single functional system tool. In addition, databases have been structured in such a way to separate the patient identities from their clinical and genetic information. Next to this, a limit can be introduced to the set of IP addresses allowed to enter a specific database. The socket server invests the quadruple role of functional storage of the servers' activity, database location, registered domains and users with their privileges; load balancing of services' requests; resources management control by adding or deleting domains and users; and authentication by way of identifying the different users' access rights.
In fact, users can simply register with the GenoMap system after which they are assigned to a domain corresponding to a database stored on a computer within the user's administrative domain. Creating a new domain involves the set-up of a database to register it with the socket server. A user group with appropriate privileges is generated with access to the services by way of the client applets. At present, GenoMap is still being applied on a rather small scale with clients contacting their paired servers individually by registering on an applet/server pair basis. Soon, the architecture will be expanded to a separate Socket Server process (SSP) that registers with the Web Server and provides automatic load-balancing. In this way, valid GenoMap users will be authenticated, whereas the location information will be passed for the applet clients needing to contact various backend servers and the Database. The GenoMap paper provides you with all the details.