MAVIS integrates with SV callers, job schedulers, and aligners. While some of these dependencies are optional, all currently supported options are detailed below. The versions column in the tables below list all the versions which were tested for each tool. Each version listed is known to be compatible with MAVIS.
MAVIS can be run locally without a job scheduler
MAVIS_SCHEDULER=LOCAL) however, due to the computational resources
generally required, it is recommended that you use one of the supported
schedulers listed below.
Users requiring support for other schedulers may make a request by submitting an issue to our github page. Additionally, developers looking to extend the functionality may submit a pull request (Please see the guidelines for contributors
MAVIS running locally uses the python
concurrent.futures library to manage
When setting the aligner you will also need to set the aligner_reference to match
MAVIS supports output from a wide-variety of SV callers. Assumptions are made for each tool based on interpretation of the output and the publications for each tool. The tools and versions currently supported are given below. Versions listed indicate the version of the tool for which output files have been tested as input into MAVIS
MAVIS also supports a general VCF input. It should be noted however that the tool tracked will only be listed as 'vcf' then.
Trans-ABySS: The trans-abyss version used was an in-house dev version. However the output columns are compatible with 1.4.8 as that was the version branched from. Additionally, although indels can be used from both genome and transcriptome outputs of Trans-ABySS, it is reccommended to only use the genome indel calls as the transcriptome indels calls (for versions tested) introduce a very high number of false positives. This will slow down validation. It is much faster to simply use the genome indels for both genome and transcriptome.
Some post-processing on the delly output files is generally done prior to input. The output BCF files are converted to a VCF file
bcftools concat -f /path/to/file/with/vcf/list --allow-overlaps --output-type v --output combined.vcf
Writing A Custom Conversion Script
Logic Example - Chimerascan
The following is a description of how the conversion script for Chimerascan was generated. While this is a built-in conversion command now, the logic could also have been put in an external script. As mentioned above, there are a number of assumptions that had to be made about the tools output to convert it to the standard mavis format. Assumptions were then verified by reviewing at a series of called events in IGV. In the current example, Chimerascan output has six columns of interest that were used in the conversion
The above columns describe two segments which are joined. MAVIS requires the position of the join. It was assumed that the segments are always joined as a sense fusion. Using this assumption there are four logical cases to determine the position of the breakpoints.
i.e. the first case would be: If both strands are positive, then the end of the five-prime segment (end5p) is the first breakpoint and the start of the three-prime segment is the second breakpoint
Calling A Custom Conversion Script
Custom conversion scripts can be specified during
automatic config generation
Any external conversion scripts must take a
-o option which requires a
single outputfile argument. This outputfile must be the converted file
output by the script. Additionally, the conversion script must be
specified by its full path name and have executeable permissions.
In the following example the user has created a custom conversion script
my_convert_script.py which they are passing an input file named
mavis config --external_conversion my_converted_input1 "my_convert_script.py my_input1.txt ... "
This will then be called during the pipeline step as
my_convert_script.py my_input1.txt ... -o /path/to/output/dir/converted_inputs/my_converted_input1.tab
You can also re-use the same conversion script if you have multiple inputs to convert simply by specifying a different alias
mavis config \ --external_conversion my_converted_input1 "my_convert_script.py my_input1.txt" \ --external_conversion my_converted_input2 "my_convert_script.py my_input2.txt"
General VCF inputs
Assuming that the tool outputting the VCF file follows standard conventions, then it is possible to use a general VCF conversion that is not tool-specific. Given the wide variety in content for VCF files, MAVIS makes a number of assumptions and the VCF conversion may not work for all VCFs. In general MAVIS follows the VCF 4.2 specification. If the input tool you are using differs, it would be better to use a custom conversion script.
Assumptions on non-standard INFO fields
PRECISEif given, Confidence intervals are ignored if given in favour of exact breakpoint calls using pos and END as the breakpoint positions
CTvalues if given are representative of the breakpoint orientations.
CHR2is given for all interchromosomal events
Translating BND type Alt fields
There are four possible configurations for the alt field of a BND type structural variant based on the VCF specification. These correspond 1-1 to the orientation types for MAVIS translocation structural variants.
r = reference base/seq u = untemplated sequence/alternate sequence p = chromosome:position