logistic regression using HLA alllelic data

Question

[monalc40]

I have a case-control dataset and I want to perform logistic regression and conditional logistic regression based on HLA multi-allelic data, using r. I want to find the effect on specific alleles on a trait. How do I do this what is the format. Most examples are based on SNP biallelic data. For instance at HLA-A I may have up to 30 unique alleles, at HLA-B it could be 50. Should I recode all the alleles and perform logistic regression on genotype pairs?

[Kevin Blighe]

If you are merely asking this as a technical question, then you can do this in R via glm(). Your SNP predictors can be encoded categorically for as AA, AB, BB, or continuously as minor allele counts.

Kevin

Answer 1

[Lemire]

Find a way to produce a data frame containing the counts of each alleles that you see, and the case-controls status. E.g. (fake data)

> df

 DX DRB1.0401 DRB1.0404 DRB1.0405 DRB1.0408
1  0         0         0         1         1
2  0         0         0         0         2
3  0         0         0         0         2
4  1         1         0         0         1
5  1         1         0         0         1
6  1         0         1         0         1

If you are interested on the effect of a specific allele, then you can do, e.g.

summary(glm( DX ~ DRB1.0401 , family="binomial", data=df ) )

If you are interested in the effect of your HLA locus as a whole, then you can do, e.g.

full<- glm( DX ~ DRB1.0401 + DRB1.0404 + DRB1.0405 + DRB1.0408, 
   family="binomial", data=df ) 
null<- glm( DX ~ 1 , family="binomial", data=df ) 

anova( null, full , test="Chisq")

adding covariates to the models if deemed necessary.

[monalc40]

The problem has been solved, thanks

[curious]

To expand on this, how would it look if you did have covariate (sex). For example say I have a multialleleic locus with three possible snps:

 data <- data.frame("snp1"=c(runif(n=150, min=0,max=2),
                          c(runif(n=50, min=0,max=2))),
                  "snp2"=c(runif(n=50, min=0, max=.2),
                           runif(n=50, min=0, max=.2),
                           runif(n=50, min=1.5, max=2),
                           runif(n=50, min=1.5, max=2)),
                  "snp3"=c(runif(n=50, min=0, max=.2),
                          runif(n=50, min=0, max=.2),
                          runif(n=50, min=1.5, max=2),
                          runif(n=50, min=1.5, max=2)),
                  "sex"=runif(n=50, min=0, max=1),
                   "disease"=c(rbinom(150, 1, 0.1),
                               rbinom(50, 1, 0.9)))

to test locus at whole I would do this:

multi_snp_full <- glm(disease ~ snp2 + snp3 + sex, data=data, family="binomial")
null <- glm(disease ~ sex, data=data, family="binomial") 
anova( null, multi_snp_full , test="Chisq")

If I wanted to go back and test snp2 specifically, would it just be this (with no LR test)?

single_snp_test <- glm(disease ~ snp2 + sex, data=data, family="binomial")