Medical research strives towards a better understanding of the molecular pathophysiology. Due to advancements in laboratory equipment, the number of measurements taken per probe in medical studies could be increased. This increase in observations per individual has to be accompanied by an increase in the number of individuals to account for the problem of multiple testing. The resulting data sets are huge and these technical diffi- culties require medical investigators to seek external advice. Exemplified by problems encountered in the statistical genomics of an immune mediated disease, this work takes a new approach on data presentation to aid in investigative biomedical research.

The key problems addressed are secure and failure-resistant storage of huge data sets and their presentation to the researcher. To aid in the investigative work, the presentation should react to user requests promptly, preferably in real-time. These goals are achieved by capitalizing on related developments in high performance computing and distributed data storage.

This web application is using the distributed database Apache Cassandra for a highly scalable and failure-resistant data storage. The front-end is implemented using Google Web Toolkit and allows the user a dynamic browsing of his data set in table, gauge and map presentation modes. Targeted users are researchers who need to cope with large-scale expression QTL experiments where common presentation and analysis frameworks fail to provide the required interactivity for investigative research.

You will need:

• Java 1.6
• Google Web Toolkit 2.0.3 (for compiling)
• shared PostgreSQL 8.4 server for management data
• one static IP address for every node
• ports 8080 and 9160 mandatory accessible peer-to-peer between nodes
• ports 7000 and 7001 optionally accessible peer-to-peer between nodes (think huge performance benefit)

First, you will need to create the folder war/WEB-INF/lib and place the following jar files into it: MD5 (war/WEB-INF/lib/JRI.jar) = 0dafd2203b9af4309b0adf1f8d4c278e MD5 (war/WEB-INF/lib/antlr-2.7.6.jar) = 97c6bb68108a3d68094eab0f67157962 MD5 (war/WEB-INF/lib/commons-collections-3.1.jar) = d1dcb0fbee884bb855bb327b8190af36 MD5 (war/WEB-INF/lib/dom4j-1.6.1.jar) = 4d8f51d3fe3900efc6e395be48030d6d MD5 (war/WEB-INF/lib/gwt-servlet.jar) = 5e873b12c7ece5194196f8da03efedbb MD5 (war/WEB-INF/lib/gwt-user.jar) = d846e1c4693d0479f1d9011f830e0cd7 MD5 (war/WEB-INF/lib/hibernate3.jar) = 84b876f227d80ed857f251800a8bc759 MD5 (war/WEB-INF/lib/javassist-3.9.0.GA.jar) = e85b97419a1bbea4c6e432b3ab1d4197 MD5 (war/WEB-INF/lib/jpa-api-2.0-cr-1.jar) = 4a28f3e4fee05e13cf45e6186bf8727e MD5 (war/WEB-INF/lib/jta-1.1.jar) = ffc9ca23acc1665c90dbbe715645564f MD5 (war/WEB-INF/lib/libthrift-r820831.jar) = d0e0d615bfd80863d050949d0e71404a MD5 (war/WEB-INF/lib/log4j-1.2.15.jar) = 4d4609998fbc124ce6f0d1d48fca2614 MD5 (war/WEB-INF/lib/org.restlet.ext.servlet.jar) = fde50e68350e227aea892f7720077176 MD5 (war/WEB-INF/lib/org.restlet.jar) = c03b90abd695143f1197c090f1b8d210 MD5 (war/WEB-INF/lib/persistence-api.jar) = aeb56ad8210370d0cd5c0e995eb0d16c MD5 (war/WEB-INF/lib/postgresql-8.4-701.jdbc4.jar) = 298a1db3ea7927d17ce18a5a49c5bbf7 MD5 (war/WEB-INF/lib/slf4j-api-1.5.8.jar) = f552fed65cab609d57f9aed8de435651 MD5 (war/WEB-INF/lib/slf4j-log4j12-1.5.8.jar) = e833f7d233d6240c919b57b51a52002e

Next, modify src/hibernate.cfg.xml and replace the USERNAME and PASSWORD placeholders with the username and password for your shared PostgreSQL database server.

Third, modify src/de/uni_luebeck/inb/krabbenhoeft/eQTL/server/processors/ConvertCMorganToBPProcessor_MMus.java

Optional: Replace "INSERT YOUR MARKER DATA HERE" with your centimorgan to basepair mapping tables in the following format:

tmp = new HashMap<Double, Integer>();
map.put("1", tmp); // chromosome 1
tmp.put(0.0, 123); // centimorgan 0.0 mappes to basepair 123
tmp.put(3.0, 456); // centimorgan 3.0 mappes to basepair 456
tmp.put(10.6, 789); // centimorgan 10.6 mappes to basepair 789

If you do not wish to use centimorgan to basepair conversions, simply remove the "INSERT YOUR MARKER DATA HERE" line.

Now you can compile the source code using Google Web Toolkit 2.0.3, which will fill the war folder with the compiled web application including server-side components.

Download Jetty 6.1.22 and extract.

Replace all contents of the webapps/war folder with the war folder you created during compilation.

Copy the hajo.xml file from deployment_helpers to contexts/hajo.xml

Start Jetty :)

If you would like to use the integrated R shell, start Jetty like this: export R_HOME=/path/to/R export JAVA_OPTIONS=" -ea -d32 -Xms128M -Xmx2G -Djava.library.path=/path/to/R/library/rJava/jri: " sh bin/jetty.sh run

Download Apache Cassandra 0.5.1 and extract.

Copy the example storage-conf.xml to conf/storage-conf.xml and modify the Seeds section. You should list the IP addresses of some nodes here. While Apache Cassandra nodes can communicate and find each other using a peer-to-peer gossip protocol, at least one IP is needed such that the newly starting node knows which cluster it belongs to.

<Seeds>
	<Seed>IP HERE</Seed>
	<Seed>ANOTHER IP HERE</Seed>
</Seeds>

Every load balancer should do. Just point it at port 8080 on each node, since that is where the Jetty server is running.

During development, nginx 0.7.65 was used for testing. Once again, downlaod and extract.

Then replace conf/nginx.conf with the example nginx.conf from deployment_helpers. You will need to edit the following section to list all node IP addresses:

upstream  mysite  {
      server   192.168.1.80:8080;
      server   192.168.1.81:8080;
      server   192.168.1.85:8080;
}

You might have noticed that Jetty needs no IP set-up and Cassandra needs only a generic set of seeding IPs which can be the same for all nodes. That is intentional :)

Setting up a new node is as easy as copying the Jetty and the Cassandra folders from one existing node and starting both.

After that, you will need to add the new node's IP to the load balancer and that's it!

(If you use nginx, simply edit the nginx.conf and add the new IP to the "upstream mysite" section. No restart required.)

The generated Thrift interfaces in the gen-java folder are licensed under Apache License 2.0.

The source code written by Hajo Nils Krabbenhöft is licensed under GNU Lesser General Public License 3.0.