Population Prediction with 1000 Genomes data using PCA and a Random Forest Classifier

This notebook contains code for predicting 1000 Genomes populations using principal component analysis (PCA) and a random forest classifier

Getting the data

The VCFs can be downloaded from here

The population information file can be downloaded from here: ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/release/20130502/integrated_call_samples_v3.20130502.ALL.panel

Pre-processing

Before using the 1000 Genomes VCF files, we first do a bit of pre-processing. First, we concatenate the VCFs together into one giant VCF file using bcftools. The command for doing this is in the script concat_phase3_vcfs.sh.

Then, we filter the variants to use a subset using vcftools. The command for doing this is:

vcftools --gzvcf all.1kg.phase3_shapeit2_mvncall_integrated_v1b.20130502.vcf.gz --snps pruned_SS2_ids_out.txt --recode

Finally, the data should be ready for processing.