HLAassoc

(1) Introduction

This tapa provides a HLA-focused statistical association analysis. Depends on the disease of focus, we can either apply a linear (quantitative trait) or logisitc (binary trait) for single variant association test. We can also perform a conditional analysis (--condition or --condition-list) for fine-mapping studies.

Due to the high LD structure within the extended MHC region, a haplotype based (OMNIBUS) approach is often implemented in the HLA fine-mapping studies. For each amino acid position, we perform a multiallelic association between trait of interest and a haplotype matrix (of the amino acid positions). To get an omnibus P-value for each position, we estimated the effect of each amino acid by assessing the significance of the improvement in fit by calculating the in-model fit, compared to a null model following an F-distribution with m-1 degrees of freedom, where m is the number of residues in an amino acid position. This is implemented using an ANOVA test in R. The most frequent haplotype was excluded from a haplotype matrix as a reference haplotype for association.

(2) Input and output

You would need the imputed vcf file and a phenotype file (--pheno) to start your association test. You can alsp provide other covariates using the --covar flag.

(3) Usage examples

HLAassoc in HLA-TAPAS has to be implemented in the directory of main project folder. (i.e. 'HLA-TAPAS/' where 'HLA-TAPAS.py' script is.)

$ cd ../ 
# Change your current directory to the HLA-TAPAS main project folder.

(a) Logistic Regression

$ python -m HLAassoc LOGISTIC \
    --vcf HLAassoc/example/IMPUTED.1958BC.bgl.phased.vcf.gz \
    --out Myassoc/IMPUTED.1958BC \
    --pheno HLAassoc/example/1958BC.phe \
    --pheno-name p1

If user wants to make HLA marker have un-transformed allele name, say if your reference panel is only at the G-group resolution, you might want to transform the 4-field alleles back to G-group alleles for association studies. Users can get this by passing '--hped' and '--chped' arguments.

$ python -m HLAassoc LOGISTIC \
    --vcf HLAassoc/example/IMPUTED.1958BC.bgl.phased.vcf.gz \
    --out Myassoc/IMPUTED.1958BC.rev_map \
    --pheno HLassoc/example/1958BC.phe \
    --pheno-name p1 \
    --hped HLAassoc/example/1958BC.Ggroup.hped \
    --chped HLAassoc/example/1958BC.imgt3320.4field.chped

(b) Linear Regression

Similar to the logisitc regression, here's how linear regression works:

$ python -m HLAassoc LOGISTIC \
    --vcf HLAassoc/example/IMPUTED.1958BC.bgl.phased.vcf.gz \
    --out Myassoc/IMPUTED.1958BC \
    --pheno HLAassoc/example/1958BC.phe \
    --pheno-name p2

(c) Omnibus Test

Omnibus Test requires next 7 files, each of which contains necessary information for Omnibus Test.

1. (Imputed) Phased BEAGLE file. (*.bgl.phased)
2. PLINK bim file of reference panel. (*.bim)
3. PLINK fam file of target samples. (*.fam)
4. Top 10 Principal Component information generated by PLINK. (*.pcs; with header.)
5. PLINK phenotype file of target samples. (*.pheno; Case: 1/ Control: 0; without header.)
6. Sex information of target samples. (*.sex; without header.)
7. Population information of target samples. (*.pop; without header.)

Assuming 1 - 6 files have the same file prefix and those specified file extensions, those 6 files can be passed to the Omnibus Test all at once with the '--file' argument.

$ python -m HLAassoc OMNIBUS \
    --file HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.IMPUTED.chr6.hg18.100+100 \
    --pop HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.IMPUTED.chr6.hg18.100+100.pop \
    --out Myassoc/Case+Control+1000G_EUR_REF.OMNIBUS \
    --aa-only \
    --maf-threshold 0

The above command is equivalent to the next command.

$ python -m HLAassoc OMNIBUS \
    --phased HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.IMPUTED.chr6.hg18.100+100.bgl.phased \
    --bim HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.IMPUTED.chr6.hg18.100+100.bim \
    --fam HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.IMPUTED.chr6.hg18.100+100.fam \
    --pcs HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.IMPUTED.chr6.hg18.100+100.pcs \
    --pheno HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.IMPUTED.chr6.hg18.100+100.pheno \
    --sex HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.IMPUTED.chr6.hg18.100+100.sex \
    --pop HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.IMPUTED.chr6.hg18.100+100.pop \
    --out Myassoc/Case+Control+1000G_EUR_REF.OMNIBUS \
    --aa-only \
    --maf-threshold 0

Or, next minimal 3 files can be passed to perform the Omnibus Test.

1. (Imputed) VCF file generated from SNP2HLA. (*.vcf.gz)
2. PLINK bim file of reference panel. (*.bim)
3. PLINK fam file of target samples. (*.fam)

Phased BEAGLE file and Top 10 PCs will be generated from the given VCF file. PLINK fam file must contain Sex and Phenotype information. All samples will be considered to be originated from same population unless the '--pop' argument is given.

$ python -m HLAassoc OMNIBUS \
    --vcf HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.IMPUTED.bgl.phased.vcf.gz \
    --bim HLAassoc/example/OMNIBUS/Case+Control+1000G_EUR_REF.REF.bglv4.bim \
    --fam HLAassoc/example/OMNIBUS/Case+Control.300+300.chr6.hg18.fam \
    --pheno HLAassoc/example/OMNIBUS/Case+Control.300+300.phe \
    --out Myassoc/Case+Control+1000G_EUR_REF.OMNIBUS \
    --aa-only \
    --maf-threshold 0