The Exomiser program was originally written purely for exome sequence analysis, however in version 7.0.0 it was extended to be able to analyse whole genome sequences too. The underlying framework and phenotype matching algorthims are identical to the previous version, although major architectural changes were introduced which enabled users to configure the exomiser to run analyses in a user-defined manner. The original command-line parameters and settings file inputs are still maintained and when used will run the original exomiser algorithm. Using the new analysis file input it is still possible to run the original exomiser algorithm (see test-analysis-exome.yml). The advantage here is that it is now possible to run an analysis over a whole-genome sample, yet treat it as an artificial exome for example. We would encourage users to use the new yml format for all their analyses as it also provides a record of exactly how a sample was analysed. This can also be easily shared with collaborators or embedded in publications. The Genomiser algorithm is contained in the test-analysis-genome.yml which has been optimised for both speed and memory usage in analysing whole genomes. Deviating from this default could significantly increase both the time and RAM used for an analysis.
Configuration
The Exomiser can be configured to run in various ways either purely through the command-line interface or by specifying various input file formats and run parameters through the settings file or with an analysis configuration file for maximum flexibility.
Command-line interface
The exomiser is typically always accessed via the command-line to some extent - it is launched using the java command:
java -jar exomiser-cli-7.2.3.jar
It is a also good practice to specify the lower and upper limits of the memory required by the Java VM using the -Xms and -Xmx switches. Typically exomiser can be run using about 8-10GB RAM and in many cases only a 2-4GB RAM, but this will vary considerably depending on the sample size and and the run parameters. Retaining all variants in memory irrespective of their PASS/FAIL state is the least efficient way to analyse a sample. For samples containing more than a few hundred thousand variants, we strongly recommend using exomiser either on a pre-filtered VCF if retaining all variants using the --full-analysis true argument or using the PASS_ONLY analysis mode in the analysis file.
Settings file
Settings files contain all the parameters passed in on the command-line so you can just point exomiser to a settings file using the --settings-file argument. See example.settings and test.settings.
java -Xms2g -Xmx4g -jar exomiser-cli-7.2.3.jar --settings-file test.settings
Alternatively you can mix up a settings file and override settings by specifying them on the command line:
java -Xms2g -Xmx4g -jar exomiser-cli-7.2.3.jar --settings-file test.settings --prioritiser=phenix
Settings can also be run in batch mode. Simply put the path to each settings file in the batch file - one file path per line.
java -Xms2g -Xmx4g -jar exomiser-cli-7.2.3.jar --batch-file batch.txt
Analysis file
Instead of specifying all commands on the command-line you can specify exomiser to use the configuration contained an analysis file using the --analysis argument.
Analysis files contain all possible options for running an analysis including the ability to specify variant frequency
and pathogenicity data sources and the ability to tweak the order that analysis steps are performed.
java -Xms2g -Xmx4g -jar exomiser-cli-7.2.3.jar --analysis test-analysis-exome.yml
These files can also be used to run full-genomes, however they will require substantially more RAM to do so. For example a 4.4 million variant analysis requires approximately 12GB RAM. However, RAM requirements can be substantially reduced by setting the analysisMode option to PASS_ONLY.
Analyses can also be run in batches using the --analysis-batch command. This requires a file containing the paths to each analysis with one path per line as input.
java -Xms4g -Xmx8g -jar exomiser-cli-7.2.3.jar --analysis-batch test-batch-analysis.txt
If you have several genomes/exomes to analyse this is highly recommended as it will remove the start-time overhead and also allow the user to make use of caches as described in the application.properties file. Used correctly, this can save a lot of time at the expense of RAM as variant frequency and pathogenicity data will be cached for the most common variants cutting down on calls to the database.