🚧 Under construction 🏗️

Haplotype Wide Association Study (HWAS) pipeline

Here is a collection of scripts that are used for haplotype wide association studies (HWAS) in the Palmer Lab. Scripts are not meant to be completely generalizable, but instead for specific use by the Palmer Lab given the Palmer Lab infrastructure.

Dependencies

• R >= 4.1.1
  • qtl2: package for association testing
• python >= 3.10
  • psycopg
• hgrm: Lab tool for computing kinship matrices
• bcftools >= 1.21
• tabix (htslib) >= 1.21
• bgzip (htslib) >= 1.21

Interactive use of hwas

hwas consists of several steps, each of which are implemented as a subcommand. To start, use the subcommand init

hwas init schema_name phenotype

which will create directories:

- schema_name/
    |
    |- phenotype/
        |
        |- 001/
            |
            |- config

The hwas tools use the written config file to specify or change parameters. The config file is divided into sections, in each section there are option and value pairs. To see what sections exists simply

hwas config

which will list the sections. To see the option and value pairs in a section

hwas config -s section_name

or to see the value of an individual section and option pair

hwas config -s section_name -o option_name

and lastly, to change the value of a section and option pair

hwas config -s section_name -o option_name -v value

A subset of requisite parameters may be specified in the analysis step, for example let's query data from a postgres data base with the Palmer Lab organization,

hwas query --dbname my_db_name \
    --host hostname \
    --port port_number \
    --db_user me

if the database requires a password you'll need to set the password in the enivronmental variable PALMER_DB_PW prior to querying the data. The query subcommand will produce two files, let's assume that the current working directory is path_to_data/schema_name/phenotype/001,

./
|- config
|- covariates.csv
|- phenotype.csv

The data queried from the database consists of all relevant samples, in which any one sample may or may not be genotyped. To find the samples with covariates, phenotype, and genotypes available run

hwas intersect path_to_vcf/genotypes.vcf

This command will overwrite the covariates.csv and phenotype.csv files with the data of samples specified in the samples file.

./
|- config
|- covariates.csv
|- phenotype.csv
|- samples

The samples file is critical, as it is repeatedly used in subsequent steps, so please do not modify or delete.

Next we can compute the genetic relationship matrix G of each sample $i$ and $j$ using the expected haplotype counts $X_{ih}$ and defined as

$$ G_{ij} = \sum_{m=1}^{M_\text{loci}} \sum_{h = 1}^{N_\text{founders}} X_{imh} X_{jmh} $$

with $h$ and $m$ being the indices over $N_\text{founders}$ ancestral founder haplotypes and $M_\text{loci}$ genetic loci. To compute this hgrm for chr12 we would use the hgrm submodule as follows

hwas hgrm chr12

This command makes the direcotry hgrm, if it doesn't already exist, and writes $G$ to file chr12.mat. This is a comma delimited text file with $N_\text{samples}$ rows and $N_\text{samples}$ columns. The order of samples is that in the samples file. Now suppose we next compute the hgrm of chr1

hwas hgrm chr1

would result in the directory structure

./
|- config
|- covariates.csv
|- phenotype.csv
|- samples
|- hgrm/
    |- chr1.mat
    |- chr12.mat

Pipeline use of hwas on hpc system with SLURM

Acknowledgment

This code has been reviewed by Claude, the AI assistant from Anthropic. The code was designed and implemented by Robert Vogel, code recommendations that were provided by Claude were adapted and implemented by Robert Vogel.