Test Data for Structure_threader

In this directory you will find the data that was used to benchmark Structure_threader.

BigTestData.str.tar.xz

This file is a fastStructure formatted input file which was used to benchmark fastStructure. This is a large SNP file (1000 SNPs across 1000 individuals) which was obtained from the 1000 genomes project. The file was downloaded from chromossome 22, and was then filtered using vcftools with the following criteria:

  • only biallelic, non-singleton SNV sites
  • SNvs must be at lest 2KB apart from each other
  • minor allele frequency < 0.05

The used command was:

./vcftools --gzvcf \
ALL.chr22.phase3_shapeit2_mvncall_integrated_v5a.20130502.genotypes.vcf.gz \
--maf 0.05 --thin 2000 --min-alleles 2 --max-alleles 2 --non-ref-ac 2 \
--recode --chr 22 --out Chr22

This was the criteria that was used on the admixture analysis of the 1000 genomes project.

The file was then converted to structure format with PGDSpider. To further reduce the dataset (for faster benchmarking), the file was then processed with cut and head and finally compressed with xz.

The used commands were:

cut -d " " -f 1-1000 Chr22.recode.str | head -n 2000 > BigTestData.str
tar cvfJ BigTestData.str.tar.xz BigTestData.str

BigTestData.bed.tar.xz

This file is a PLINK formatted .bed, .bim and .fam set of files. They were obtained in the exact same way as BigTestData.str.tar.xz, except for the conversion using PGDSPIDER, which was not used. Instead, the filtered VCF file was reduced to 501 individuals and 1000 SNPs with the following command:

head -n 1253 Chr22.recode.vcf |cut -f 1-510 > Testdata.vcf

This file was then converted to the PLINK format and compressed with the following commands:

plink1.9 --vcf Testdata.vcf
mv plink.bed BigTestData.bed
mv plink.fam BigTestData.fam
mv plink.bim BigTestData.bim
tar cvfJ BigTestData.bed.tar.xz BigTestData.bed BigTestData.fam BigTestData.bim

BigTestData.vcf.tar.xz

This file is *VCF* formatted. It was obtained in the exact same way as `BigTestData.str.tar.xz`, except for the conversion using *PGDSPIDER*, which was not used. Instead, the filtered VCF file was reduced to 501 individuals and 1000 SNPs and compressed with the following command:

    head -n 1253 Chr22.recode.vcf |cut -f 1-510 > BigTestData.vcf
    tar cvfJ BigTestData.vcf.tar.xz BigTestData.vcf

extraparams and mainparams

The STRUCTURE parameter files that were used in the benchmarking process.

joblist.txt

The joblist used to benchmark ParallelStructure. Consists of 16 jobs, 4 values of "K" with 4 replicates each.

SmallTestData.structure

This file is a Structure formatted input file which was used to benchmark STRUCTURE and MavericK. This is a medium sized SNP file (80 SNPs) which was obtained from the 1000 genomes project. The file was downloaded from chromossome 22, and was then filtered using vcftools following the same criteria and commands as the BigTestData.str file.

The used commands were:

cut -d " " -f 1-80 SmallData.structure > SmallData302SNPs.structure
head -n 201 SmallData302SNPs.structure > SmallTestData.structure

parameter.txt

The *MavericK* parameter file that is used in the unit tests.

mav_benchmark_parameters

The file with the *MAvericK* benchmark parameters.