biotdg: Bioinformatics Test Data Generator

biotdg can generate mutations based on vcf files for genomes where the chromosomes have different ploidy. It was made to create test genomes for pipelines that correctly handle the ploidy of sex chromosomes. It can also be used to create test data for pipelines that handle triploid species, such as banana, or for pipelines that discover chromosome imbalances, such as trisomy-21 (Down syndrome) and XXY males (Klinefelter syndrome).

biotdg uses a reference genome, a ploidy table and a vcf file to create a "true genome" for a sample. For example, if the ploidy table states that chr21 has a ploidy of 3 then the "true genome" will have three copies of chr21. Each chr21 copy will have its own mutations based on the vcf file.

After creating the "true genome" fasta file. biotdg uses the dwgsim program to generate fastq reads.

Usage

usage: biotdg [-h] [--version] -r REFERENCE --vcf VCF -p PLOIDY_TABLE -s
              SAMPLE_NAME [-z RANDOM_SEED] [-l READ_LENGTH] [-C COVERAGE]
              [-e READ1_ERROR_RATE] [-E READ2_ERROR_RATE]
              [-n MAXIMUM_N_NUMBER] [-o OUTPUT_DIR]

Bioinformatics Test Data Generator

optional arguments:
  -h, --help            show this help message and exit
  --version             show program's version number and exit
  -r REFERENCE, --reference REFERENCE
                        Reference genome for the sample.
  --vcf VCF             VCF file with mutations.
  -p PLOIDY_TABLE, --ploidy-table PLOIDY_TABLE
                        Tab-delimited file with two columns specifying the
                        chromosome name and its ploidy. By default all
                        chromosomes have a ploidy of 2.
  -s SAMPLE_NAME, --sample-name SAMPLE_NAME
                        Name of the sample to generate. The sample must be in
                        the VCF file.
  -z RANDOM_SEED, --random-seed RANDOM_SEED
                        Random seed for dwgsim (default: 1).
  -l READ_LENGTH, --read-length READ_LENGTH
                        Read length to be used by dwgsim.
  -C COVERAGE, --coverage COVERAGE
                        Average coverage for the generated reads. NOTE: This
                        is multiplied by the ploidy of the chromosome.
  -e READ1_ERROR_RATE, --read1-error-rate READ1_ERROR_RATE
                        Same as -e flag in dwgsim. per base/color/flow error
                        rate of the first read.
  -E READ2_ERROR_RATE, --read2-error-rate READ2_ERROR_RATE
                        Same as -E flag in dwgsim. per base/color/flow error
                        rate of the second read.
  -n MAXIMUM_N_NUMBER, --maximum-n-number MAXIMUM_N_NUMBER
                        Maximum number of Ns allowed in a given read.
  -o OUTPUT_DIR, --output-dir OUTPUT_DIR

Example

Given the following reference.fasta file

>chr1
GATTACA
GATTACA
GATTACA
>chrX
AGTCAGTCAGTC
>chrY
AGAATC

the following ploidy table.tsv

chr1        3
chrX        2
chrY        1

and the following vcf:

##fileformat=VCFv4.1
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##contig=<ID=chr1,length=21>
##contig=<ID=chrX,length=12>
##contig=<ID=chrY,length=6>
#CHROM      POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  sample1
chr1        4       .       T       A,C,G   .       .       .       GT      1/2/3
chr1        7       .       A       T       .       .       .       GT      0/1/0
chrX        1       .       A       T       .       .       .       GT      0/1
chrX        2       .       G       T       .       .       .       GT      0/0
chrY        4       .       A       C       .       .       .       GT      1

A "true genome" for sample1 looks like this:

>chr1_0
GATAACAGATTACAGATTACA
>chr1_1
GATCACTGATTACAGATTACA
>chr1_2
GATGACAGATTACAGATTACA
>chrX_0
AGTCAGTCAGTC
>chrX_1
TGTCAGTCAGTC
>chrY_0
AGACTC

Note

Mutations are always generated in a phased manner. A _0 chromosome will receive all the genotypes in the VCF that are at index 0 (the outer left one). This is true even if the variants are not described as phased in the vcf.

Why biotdg and not dwgsim?

dwgsim has excellent capabilities for generating reads that are close to real data. Therefore dwgsim is used by biotdg in this capacity.

dwgsim can also generate mutations randomly and output these in VCF format. It also has the capability to use a VCF to generate mutations. This VCF-based method was not deemed sufficient for the following reasons:

• Very poorly documented.
• Only allows ploidy of 1 or 2. There is an option '3' but that does something different.
• How exactly mutations are generated is unknown. Is it aware of phasing? If so, how does it handle it?

biotdg handles the creation of the "true genome" transparently and then uses dwgsim to generate reads. biotdg can handle genomes with mixed ploidies (as is the case for most species with a sex chromosome) well.

Known limitations

• Overlapping mutations are not handled properly. (Probably not a concern for generating test data.)
• Mutations are always generated in a phased manner. This was easier to implement than an unphased manner. It is also more transparent. Some extra work will be required to handle unphased generation of mutations.
• biotdg is only tested with SNPs. Indels and other variant types were not tested.