remote.sh is written to be run on the master node of a cluster running a custom image of CentOS6 here.
After building the machine image and pushing to AWS, spin up a cluster using the AMI.
metanetworkSynapse requires the openmpi-x86_64 module on the cluster.
To generate networks using the expression matrix located at syn1234567, first edit config.sh so that dataSynId = syn1234567, then run the localhost script (this suggests, but does not require, a .synapseConfig file in your home [~] directory for pushing the networks to Synapse from the master node):
sh metanetworkSynapse/localhost.sh /path/to/private/key.pem your.master.nodes.public.DNS https://github.com/a/version/of/metanetworkSynapse.giti.e., the call to localhost.sh might look like sh localhost.sh ~/.aws/myKey.pem ec2-52-55-94-233.compute-1.amazonaws.com https://github.com/blogsdon/metanetworkSynapse.git if you have a cluster running in the AWS cloud and want to use the original metanetworkSynapse repository.
https://github.com/a/version/of/metanetworkSynapse.git is the fork, branch, etc. of metanetworkSynapse you would like to run on the cluster. Your local copy of config.sh will be copied over the preexisting config.sh in that repository on the master node.
This will install all dependencies in /shared/ on your cluster and submit the network jobs to the SGE queue. Networks built via the submission.sh call are high CPU intensive and medium memory intensive jobs. Be sure you have enough cores among your compute nodes to run the most demanding jobs (# compute nodes * # CPU per compute node >= nthreadsHeavy in submission.sh).
Once the jobs are finished running, if you wish to push the networks to Synapse you must:
/shared/metanetworkSynapse/config.sh so that parentId points to a Folder on Synapse that you have write permission for.sh pushNetworks.sh [githubAPIToken]. The token needs the public_repo scope, which gives you the ability to access public repositories.Once the regular networks have been pushed to Synapse, we can build the rank consensus network.
sh submissionConsensus.shBuilding the rank consensus network is a low CPU intensive and high memory intensive job. I recommend using an instance with at least 50 times as much RAM as the size of the largest network generated via submission.sh.
Once the job submitted by submissionConsensus.sh finishes, we can push the rank consensus network to Synapse with sh pushConsensus.sh [githubAPIToken].
When running jobs, error and output logs are kept in /shared/network/errorLogs/ and /shared/network/outLogs/, respectively.
If there are other problems submitting jobs to the SGE queue, building the networks, or pushing to Synapse, we provide a small test network for debugging. The process for running the test network is the same as above, except the main configuration file is kept in configTest.sh, jobs are submitted with sh submissionTest.sh, completed networks are pushed to Synapse with sh pushTest.sh, and the rank consensus network is built with sh submissionTestConsensus.sh. Error and output logs for the test network are in /shared/testNetwork/errorLogs/ and /shared/testNetwork/outLogs/, respectively.