I would be thankful to you if you can help with the statistical approach for 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.


  1. The sample size is small.
  2. Maybe structural equation modelling could be used to find CpG sites associated with arsenic exposure and disease condition as I have very small size samples
  3. R is a preferred language.
  • 1
    $\begingroup$ This forum is not dedicated to general help requests and your post may be closed. Ask a precise bioinformatics question if you are encountering difficulties in your analysis, there is a lot of tutorials available on the internet. $\endgroup$
    – Basti
    Commented Nov 16, 2023 at 12:37
  • $\begingroup$ What computer language are you going to utilize in this project? Are you planning on an ML/AI approach? That's a major question to ask yourself. If you are planning on a statistical approach, I highly recommend R and then Python. If your skills are advanced, Id try an AI/ML approach! What databases/sources did you use to get the phenotype files in .csv format? What data types are the csv values in? Are they integers, decimals/floats, strings? Should you consider a .tsv files format instead? What's the purpose to your research question to arsenic and DNA methylation? $\endgroup$ Commented Nov 16, 2023 at 14:10
  • 1
    $\begingroup$ The question is part of a previous question here bioinformatics.stackexchange.com/questions/21823/… . If they are taken together, plus comments it's answerable. I will make the appropriate edits shortly. $\endgroup$
    – M__
    Commented Nov 16, 2023 at 15:04

1 Answer 1


Having given some thought to this, I'd still use machine learning. I would attempt to augment the negative controls however. Is there a suitable published data set? R program is caret but it high barrier to entry without appropriate support.

250 samples is just about enough. If you can augment that to 350 with negatives ... should be enough. What you do is dump all your data ("features") into a single record, including the methylation data and use toxic element exposure as the training target. The output you want is called "feature selection", what you are wanting is methylation sites to carry more "weight" than any other feature. Then you start other analytics to see what it's associating with. Its not trivial getting it working and a sceptic would complain about the lack of significance in the results, but its a good starting point.

GWAS might work, just considering methylation vs toxic element exposure. SEM, I've never used it so I don't know.


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.