Reference Input Files

There are several reference files that are required for full functionality of the MAVIS pipeline. If the same reference file will be reused often then the user may find it helpful to set reasonable defaults. Default values for any of the reference file arguments can be configured through environment variables.

To improve the install experience for the users, different configurations of the MAVIS annotations file have been made available. These files can be downloaded below, or if the required configuration is not available, instructions on generating the annotations file can be found below.

File Name (Type/Format)

Environment Variable


reference genome (fasta)



annotations (JSON)


GRCh37/Hg19 + Ensembl69
GRCh38 + Ensembl79

masking (text/tabbed)



template metadata (text/tabbed)



DGV annotations (text/tabbed)



aligner reference


GRCh37/Hg19 2bit (blat)
GRCh38 2bit (blat)

If the environment variables above are set they will be used as the default values when any step of the pipeline script is called (including generating the template config file)

Reference Genome

These are the sequence files in fasta format that are used in aligning and generating the fusion sequences.

Template Metadata

This is the file which contains the band information for the chromosomes. This is only used during visualization.

The structure of the file should look something like this

chr1    0       2300000 p36.33  gneg
chr1    2300000 5400000 p36.32  gpos25
chr1    5400000 7200000 p36.31  gneg
chr1    7200000 9200000 p36.23  gpos25
chr1    9200000 12700000        p36.22  gneg

Masking File

The masking file is a tab delimited file which contains regions that we should ignore calls in. This can be used to filter out regions with known false positives, bad mapping, centromeres, telomeres etc. An example of the expected format is shown below. The file should have four columns: chr, start, end and name.

#chr    start   end     name
chr1    0       2300000 centromere
chr1    9200000 12700000        telomere

The pre-built masking files in the downloads table above are telomere regions, centromere regions (based on the cytoband file), and nspan regions (computed with tools/

Masking is not required (can provide a header-only file), but is recommended as it will improve performance and specificity.


This is a custom file format. It is a JSON file which contains the gene, transcript, exon, translation and protein domain positional information

Pre-built annotation files can be downloaded above. The ‘best transcript’ flag is based on an in-house model. We have also pre-built the ensembl annotations file including non-coding transcripts below.


It is worth noting that using the reference annotation file including the non-coding genes will require an increase in the default amount of memory for the annotation step due to the increased size of the annotations file. On our standard COLO829 we increased the default memory for the annotation step from 12G to 18G.

GRCh37/Hg19 + Ensembl69 (includes non-coding genes)


the load_reference_genes() will only load valid translations. If the cds sequence in the annotation is not a multiple of CODON_SIZE or if a reference genome (sequences) is given and the cds start and end are not M and * amino acids as expected the translation is not loaded

Example of the JSON file structure can be seen below

        "name": string,
        "start": int,
        "end": int
        "aliases": [string, string, ...],
        "transcripts": [
                "name": string,
                "start": int,
                "end": int,
                "exons": [
                    {"start": int, "end": int, "name": string},
                "cdna_coding_start": int,
                "cdna_coding_end": int,
                "domains": [
                        "name": string,
                        "regions": [
                            {"start" aa_start, "end": aa_end}
                        "desc": string

The provided files were generated with Ensembl, however it can be generated from any database with the necessary information so long as the above JSON structure is respected.

Generating the Annotations from Ensembl

There is a helper script included with mavis to facilitate generating the custom annotations file from an instance of the Ensembl database. This uses the Ensembl perl api to connect and pull information from the database. This has been tested with both Ensembl69 and Ensembl79.

Instructions for downloading and installing the perl api can be found on the ensembl site

  1. Make sure the ensembl perl api modules are added to the PERL5LIB environment variable

Also ensure that the tools directory is on the PERL5LIB path so that the module can be found

# include tools/ module
export PERL5LIB
  1. Run the perl script

The below instructions are shown running from inside the tools directory to avoid prefixing the script name, but it is not required to be run from here provided the above step has been executed correctly.

you can view the help menu by running


you can override the default parameters (based on hard-coded defaults or environment variable content) by providing arguments to the script itself

perl --best_transcript_file /path/to/best/transcripts/file --output /path/to/output/json/file.json

or if you have configured the environment variables as given in step 2, then simply provide the output path

perl --output /path/to/output/json/file.json

DGV (Database of Genomic Variants) Annotations

The DGV annotations file contains regions corresponding to what is found in the database of genomic variants. This is used to annotate events that are found in healthy control samples and therefore may not be of interest if looking for somatic events.

The above (downloads table) files were generated from from DGV and reformatted to have 4 columns after download. We used awk to convert the raw file

awk '{print $2"\t"$3"\t"$4"\t"$1} GRCh37_hg19_variants_2016-05-15.txt >

Note in hg19 the column is called “name” and in hg38 the column is called “variantaccession”. An example is shown below

#chr     start   end     name
1       1       2300000 nsv482937
1       10001   22118   dgv1n82
1       10001   127330  nsv7879

Aligner Reference

The aligner reference file is the reference genome file used by the aligner during the validate stage. For example, if blat is the aligner then this will be a 2bit file.

MAVIS standard input file format

These requirements pertain to the columns of input files from the various tools you want to merge. The input files should be tab-delimited text files. Comments at the top of may be included. Comments should begin with two hash marks. They will be ignored when the file is read

## This is a comment

The header row contains the column names and is the first row following the comments (or the first row if no comments are included). Optionally the header row may (or may not) begin with a hash which will be stripped out on read

## This is a comment
## this is another comment
# this is the header row

A simple input file might look as follows

## File created at: 2018-01-02
## Generated by: MAVIS v1.0.0
#break1_chromosome  break1_position_start   break1_position_end break2_chromosome break2_position_start break2_position_end
X   1234    1234    X   77965   77965

Required Columns

Optional Columns

Optional Columns that are not given as input will be added with default (or command line parameter options) during the clustering stage of MAVIS as some are required for subsequent pipeline steps

  • break1_strand (defaults to not-specified during clustering)

  • break1_orientation (expanded to all possible values during clustering)

  • break2_strand (defaults to not-specified during clustering)

  • break2_orientation (expanded to all possible values during clustering)

  • opposing_strands (expanded to all possible values during clustering)

  • stranded (defaults to False during clustering)

  • library (defaults to command line library parameter during clustering)

  • protocol (defaults to command line protocol parameter during clustering)

  • tools (defaults to an empty string during clustering)

Summary by Pipeline Step

The different pipeline steps of MAVIS have different input column requirements. These are summarized below (for the pipeline steps which can act as the pipeline start)

column name




















Some native tool outputs are supported and have built in methods to convert to the above format. Any unsupported tools can be used as long as the user converts the tools native output to match the above format.