Hello,

I am a beginner in methylation analysis and I want to perform a comprehensive study integrating multiple DNA methylation datasets from GEO. I have already downloaded the raw IDAT files and series matrix files for each dataset.

So far, for a single dataset, my workflow is:

Read IDATs and extract sample metadata

Perform quality control (detection p-values, sample filtering)

Normalize data using preprocessSWAN or preprocessFunnorm

Extract M-values and beta-values

Fit univariate and multivariate models with limma

Perform downstream analyses (volcano plots, heatmaps, bump hunting, etc.)

My question is: at which step is it appropriate to integrate multiple datasets?

Should I merge raw IDATs before normalization?

Should I normalize each dataset separately and then integrate?

Are there recommended batch-correction or harmonization steps after merging?

Any guidance or references for best practices in multi-dataset methylation integration would be very helpful.

Thank you!