PyPGx

Table of Contents

• Introduction
• Dependencies
• Installation
• Stargazer
• Sun Grid Engine (SGE)
• SNP Callers
• Running in Command Line
• Running within Python

Introduction

PyPGx is a Python package for pharmacogenomics (PGx) research, which can be used as a standalone program and as a Python module. Documentation is available at Read the Docs.

Dependencies

PyPGx requires Python 3 and the following Python packages:

requests>=2
pandas>=1.0.0
bs4>=0.0.1
lxml>=4.5.0
pysam>=0.16.0
vcfgo>=0.0.10

Installation

The easiest way to install PyPGx and all of its dependencies is to use pip:

$ pip install pypgx

Stargazer

For genotype analyses PyPGx relies on Stargazer, a bioinformatics tool for calling star alleles (haplotypes) in PGx genes using data from next-generation sequencing (NGS) or single nucleotide polymorphism (SNP) array. Therefore, Stargazer must be pre-installed in order to run PyPGx commands such as bam2gt. For more information on Stargazer, please visit their official webpage and Github repository.

Sun Grid Engine (SGE)

Many PyPGx commands such as bam2gt2 rely on the Sun Grid Engine (SGE) cluster to distribute their tasks across multiple machines for speed. These commands are indicated by [SGE] and will generate a shell script, which can be run like this:

$ sh example-qsub.sh

SNP Callers

One major input for the Stargzer program is a Variant Call Format (VCF) file, which is a standard file format for storing SNP calls. Currently, PyPGx relies on two SNP callers to make VCF files: Genome Analysis Toolkit (GATK) and BCFtools. When running PyPGx commands like bam2vcf, you can pick which SNP calling algorithm to use; it is assumed that you already installed the corresponding SNP caller.

Generally speaking, GATK is considered more accurate but much slower than BCFtools. For instance, without the use of the SGE cluster, SNP calling for 70 WGS samples for the CYP2D6 gene takes 19 min to complete with GATK, but only 2 min with BCFtools. Therefore, if you have many samples and you do not have access to SGE for running parallel jobs, BCFtools may be a better choice. Of course, if you have SGE in your sever, then GATK is strongly recommended.

For more information on the SNP callers, please visit the GATK website and the BCFtools website.

Running in Command Line

For getting help:

$ pypgx -h
usage: pypgx [-h] [-v] tool ...

positional arguments:
  tool           name of the tool
    bam2gt       convert BAM files to a genotype file
    bam2gt2      convert BAM files to genotype files [SGE]
    gt2pt        convert a genotype file to phenotypes
    bam2vcf      convert BAM files to a VCF file
    bam2vcf2     convert BAM files to a VCF file [SGE]
    bam2gdf      convert BAM files to a GDF file
    gt2html      convert a genotype file to an HTML report
    bam2html     convert a BAM file to an HTML report [SGE]
    fq2bam       convert FASTQ files to BAM files [SGE]
    bam2bam      realign BAM files to another reference genome [SGE]
    bam2sdf      convert BAM files to a SDF file
    sdf2gdf      convert a SDF file to a GDF file
    pgkb         extract CPIC guidelines using PharmGKB API
    minivcf      slice VCF file
    mergevcf     merge VCF files
    summary      create summary file using Stargazer data
    meta         create meta file from summary files
    compare      compare genotype files
    check        check table files for Stargazer
    liftover     convert variants in SNP table from hg19 to hg38
    peek         find all possible star alleles from VCF file
    viewsnp      view SNP data for pairs of sample/star allele
    compgt       compute the concordance between two genotype files
    compvcf      compute the concordance between two VCF files
    unicov       compute the uniformity of sequencing coverage

optional arguments:
  -h, --help     show this help message and exit
  -v, --version  print the PyPGx version number and exit

For getting tool-specific help:

$ pypgx bam2gdf -h
usage: pypgx bam2gdf [-h] [--bam_dir DIR] [--bam_list FILE]
                     genome_build target_gene control_gene output_file
                     [bam_file [bam_file ...]]

positional arguments:
  genome_build     genome build ('hg19' or 'hg38')
  target_gene      name of target gene (e.g. 'cyp2d6')
  control_gene     name or region of control gene (e.g. ‘vdr’,
                   ‘chr12:48232319-48301814’)
  output_file      write output to this file
  bam_file         input BAM files

optional arguments:
  -h, --help       show this help message and exit
  --bam_dir DIR    treat any BAM files in DIR as input
  --bam_list FILE  read BAM files from FILE, one file path per line

For running in command line:

$ pypgx bam2gdf hg19 cyp2d6 vdr out.gdf in1.bam in2.bam

The output GDF file will look like:

Locus       Total_Depth     Average_Depth_sample    Depth_for_S1    Depth_for_S2
...
chr22:42539471      190     95      53      137
chr22:42539472      192     96      54      138
chr22:42539473      190     95      53      137
...

Running within Python

For running within Python:

from pypgx.phenotyper import phenotyper
phenotyper("cyp2d6", "*1", "*1")
phenotyper("cyp2d6", "*1", "*4")
phenotyper("cyp2d6", "*1", "*2x2")  # *2x2 is gene duplication.
phenotyper("cyp2d6", "*4", "*5")    # *5 is gene deletion.

To give:

'normal_metabolizer'
'intermediate_metabolizer'
'ultrarapid_metabolizer'
'poor_metabolizer'