Simulating multiple random genomes with VarSim using real VCF input

[sdws1983]

Hello,

I’m trying to use VarSim to simulate SNPs and SVs in a plant genome (e.g., Arabidopsis thaliana). I have several VCF files from real population data that include both SNPs and SVs, and I want to simulate dozens of random genomes using these VCFs.

However, when I use these VCF files as input to VarSim, the simulated genomes are always exactly the same across runs. I don’t understand why this happens — I was expecting some level of random sampling or stochasticity when generating each simulated genome.

Here is the command I used:

varsim.py --id ath_${i} --seed $i --simulator_executable ~/software/varsim/opt/ART/art_bin_VanillaIceCream/art_illumina \
--reference ~/reference/ath/upload_vcf/Col-PEK.genome.fasta --sv_num_ins 5000 --sv_num_del 5000 --sv_num_dup 2500 --sv_num_inv 2500 \
--vc_num_snp 500000 --vc_num_ins 20000 --vc_num_del 20000 --vc_num_mnp 5000 --vc_num_complex 2500 \
--vc_min_length_lim 0 --vc_max_length_lim 49 --sv_min_length_lim 50 --sv_max_length_lim 1000000 \
--disable_sim --vc_prop_het 0.6 --sv_prop_het 0.6 --vcfs ~/reference/ath/upload_vcf/72accs.col_pek.* \
--disable_rand_vcf --disable_rand_dgv --out_dir out_sim_${i} --log_dir log_sim_${i} --work_dir work_sim_${i}

Thanks!

[yunfeiguo]

Hi @sdws1983 please remove --disable_rand_vcf option as it prevents VarSim from sampling variants. Also you can use --vc_in_vcf option to supply pool of variants to sample from.

[sdws1983]

Thank you for your response. I removed the --disable_rand_vcf option as you suggested, but encountered the following error:

Traceback (most recent call last):
  File "/public2/home/off_huangyumin/software/varsim/varsim.py", line 721, in <module>
    java=args.java)
  File "/public2/home/off_huangyumin/software/varsim/varsim.py", line 330, in varsim_main
    raise ValueError("Sampling VCF missing")
ValueError: Sampling VCF missing

To resolve this, I replaced the --vcfs parameter with --vc_in_vcf, and the simulation ran successfully. However, I noticed that the sampled variants were only in the 1–50 bp range and did not include any structural variants (SVs).

As shown in the parameters I used earlier, I intended to simulate both small variants and SVs with different settings. Unfortunately, I don't have the additional input files required for SV simulation (e.g., DGV file or insertion sequences); I only have VCF files.

Do you have any suggestions for simulating both SNPs and SVs using just a VCF file?

[yunfeiguo]

Does your VCF have SVs? If no, you can use DGV (see https://github.com/bioinform/varsim/blob/master/tests/quickstart_test/quickstart.sh for an example).
If yes, you can customize RandVCF2VCF settings including --vc_num_ins, --vc_num_del, --vc_min_length_lim, --vc_max_length_lim etc to sample desired SVs plus SNPs.