Comparison of computation time of three virus integration-detecting methods on whole genome sequencing (WGS) dataa.

aThe computation time of the three methods on these samples were analyzed on Vanderbilt Advanced Computing Center for Research & Education (ACCRE, http://www.accre.vanderbilt.edu/), with the same configuration of CPUs in each node.bViralFusionSeq did not terminate successfully on sample 180N.cWe attempted to run VirusSeq three times on these WGS samples. The first trial failed because the size of its intermediate files exceeded our cluster quota. After getting more space, we reran VirusSeq. After non-stop running for a whole week, all our jobs were killed in server due to their exceeding allocated time – not realizing initially the long computation time of VirusSeq on WGS samples. In our latest trial of VirusSeq on February 13, 2013, we requested 35 GB memory, 8 CPUs, 30 days for each job and resubmitted our jobs to ACCRE. Seven jobs were scheduled to run first. After twelve and a half day, all these jobs were killed due to an unexpected internal network outage of ACCRE. Though we were not able to make VirusSeq terminate successfully on these WGS samples due to expensive computing, we may conclude from the data that VirusFinder runs much faster than VirusSeq.