1
$\begingroup$

I would be thankful to you if you can help with how I can use MR from my case-control study to link DNA methylation (epic array) with toxic element exposure (arsenic) and health outcome (Cardiovascular diseases).

I do have data from 125 Cardiovascular diseases and 125 non-Cardiovascular diseases individuals matched with age, sex, and smoking.I do have phenotypes files in a csv format having sex, age, smoking, disease, samples id, iAs (arsenic) values in log format and then a separate files for methylation data. we also have SNP data for these samples.

Many thanks,

$\endgroup$
4
  • $\begingroup$ Please add more detail to this question. What format are your datasets in? Can you provide any examples of that format? $\endgroup$
    – gringer
    Nov 8, 2023 at 5:08
  • $\begingroup$ added other informations, hope this is ok now. $\endgroup$ Nov 8, 2023 at 12:17
  • $\begingroup$ csv and log formats are not well specified. Please show some examples of inputs and expected outputs. $\endgroup$
    – gringer
    Nov 8, 2023 at 20:31
  • $\begingroup$ thanks gringer, I don't have whole data yet but thinking how to do this analysis step by step. $\endgroup$ Nov 13, 2023 at 9:59

1 Answer 1

1
$\begingroup$

I understand linkage disequilibrium (LD) versus panmixia, which the technical term for Mendelian randomisation and for humans PLINK is a population package. The best outcome is a correlation between LD and cardiovascular disease. It is however an unlikely outcome.

This really is a machine learning question and you would need to understand "feature selection" and will require very high accuracy scores >0.95 - however you could triple the size of our control group quite easily (recommended). This is a separate question and I'm happy to answer this there.


Just to address the comments. Inheritance probability, if that is the intended statistics, I personally don't think it would work because a 125 sample is low and the patients would need to be in cohorts of familial generations. LD is unlikely the explanation with humans but other non-humans species that might work (selection advantage).

My personal view is this is a straight machine learning calculation, but it's a separate question. If your sample is heavily stratified across each familial generation then Hardy-Weinberg might work, but my experience of patient cohorts is that is really unusual and there could still be confounders. Supervised learning (subset of ML) is 100% the way forward.

.... possibly use GWAS, but you've multiple features/training targets. GWAS is usually when there's one phenotype, ML would account for interaction in a way GWAS could not (it couldn't account for non-sequence features, i.e. mixed data).

$\endgroup$
7
  • $\begingroup$ Thanks, I would like to know what step by step analysis I need to perform here; like first I can look for arsenic exposure and CpG association using linear regression and then I can also look for CpG and health outcome association and then for common CpG site for arsenic exposure and health outcome I need to take for identify with the genetics variants(IVs) that affect the methylation at these site. And finally not sure if this is the correct approach and then what I need as input for MR. $\endgroup$ Nov 13, 2023 at 6:45
  • $\begingroup$ Thanks @bioinfonext Upvotes and accepts are more appreciated than thanks. I've answered above. An alternative approach e.g. via ML is a separate question. $\endgroup$
    – M__
    Nov 13, 2023 at 13:23
  • $\begingroup$ Thanks again, still I am not clear which R packages should I use for these analysis. $\endgroup$ Nov 13, 2023 at 15:39
  • 1
    $\begingroup$ GWAS: statgenGWAS package; machine learning its caret. LD is about PLINK and other autonomous packages. If you want a single package for an easy answer then shoehorn GWAS $\endgroup$
    – M__
    Nov 13, 2023 at 16:13
  • 1
    $\begingroup$ Or do you think here I should use structural equation modelling to find CpG sites associated with arsenic exposure and disease condition as I have very small size samples and you already point out that it may not useful for MR analysis. $\endgroup$ Nov 14, 2023 at 18:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.