limma output for epic array methylation data
1
0
Entering edit mode
Jitendra ▴ 10
@nabiyogesh-11718
Last seen 3 hours ago
United Kingdom

Hi all,

I was wondering if you could help me to understand limma linear regression output for continuous variables.

I got result for epic array methylation data and eGFR association in this format;

                 logFC       AveExpr.    t         P.Value  adj.P.Val   B
cg03546163  0.41018776  -0.8013218  7.85104192  1.13E-12    8.65E-07    12.3051291
cg17944885  -0.4884859  2.31974279  -5.7319645  6.33E-08    0.02431349  5.15237109


Could you please help me to understand what is difference between B, logFC, t, and AveExpr. and what are more important output need to consider to for eGFR and CpG association here?

I have used this below model at adjust for covariates;

#model matrix

var<-model.matrix(~logeGFR + as.factor(Gender) + Age +CD8T +CD4T +NK + Bcell +Mono ,data=targets2)


Thanks again for your all help. Please share any more documents or papers that can help to understand it more.

Many thanks

2
Entering edit mode
@james-w-macdonald-5106
Last seen 29 minutes ago
United States

If you are interpreting a continuous covariate, the value is the change in methylation for each unit increase in the covariate. In other words, for cg03546163, you have an increase in methylation of 0.41 for every unit increase in logeGPR.

Assuming that 'logeGPR' means log_2(eGPR) (it should!), every unit increase in logeGPR represents a doubling of eGPR exposure (or whatever eGPR is). Also, assuming that you are using M-values for the analysis (again, you should!), those values are also log_2 variates, so to interpret in the context of your experiment, note that 2^0.41 is 1.33, so you could interpret that coefficient to mean that the methylation at that site increases about 33% for every doubling of eGPR exposure.

0
Entering edit mode

Thanks James for help and time.

I have converted eGFR into log value by using below equation and used M value for methylation data.

#convert eGFR into natural log

targets$logeGFR=log(targets$eGFR)


Could you also help me to understand other output options like what is B and t and AveExpr here? Do they have any significance while reporting results? Thanks again!

2
Entering edit mode

Ideally you would use log base 2, as it's more easily interpreted.

All the columns are described in detail in the help page for topTable, so you can consult that. I only provided interpretation for logFC because the interpretation is different for a continuous covariate than for a factor (which is how it's described in the help page, because that's like 99% of the use-cases).

1
Entering edit mode

As well as the topTable() help page, another document is Chapter 13 of the limma User's Guide.

0
Entering edit mode

Thank you so much for all help.

Could anyone please help me understand the negative association between methylation and eGFR? Additionally, I'd like to know how to identify the CpG sites where methylation increased while eGFR values decreased, as well as the reverse scenario.

Thanks again!