Cluster mode benchmark run

We have ran InterProScan 5 (I5) in CLUSTER mode against a complete Escherichia coli proteome to give you some benchmark figures in terms of analysis runtime.

This documentation could be seen as a reference point for runtime, but also on how to set up I5 appropriate for speed improvement.

Benchmark run setup

Which version of InterProScan 5 (I5) was used for this run?


How was the set of input sequences assembled for this run?

For this run we decide to run I5 against the complete proteome of Escherichia coli (Taxon 83333). We’ve downloaded the proteome from the Reference proteomes website (RELEASE 2014_04).

RELEASE 2014_04 is based on UniProt Release 2014_04, Ensembl release 75 and Ensembl Genome release 21. The E.coli proteome for this release contains 4303 protein sequences. You can download the sequence file here.

Which I5 command was used for this run?

We switched off the pre-calculated match lookup service and turned on the CLUSTER mode.

-i 83333.fasta
-f tsv,html
-mode cluster
-clusterrunid benchmark-5.7-48.0

How does the file look like?

The default settings are very conservative. To speed up the CLUSTER mode we’ve added or changed the values of the following 6 attributes within the DEFAULT file:


The full setting file can be found here.

On which cluster/farm did we run I5?

We ran I5 on our internal LSF cluster. As of 1st of May 2014, there are approximatley 680 nodes comprising 16,000 hyper-threaded CPU cores.

Benchmark run outcome

| | Run 1 | Run 2 | Run 3 | |:|:———-|:———-|:———-| |Wall clock time|1h 37min | N/A | N/A | |Max number of workers| 45 | N/A | N/A |