Hi all,

I have a complex dataset that I'm looking for the right tool for analysis. I think limma might have the capabilities I need, but I'm not sure how to appropriately model this.

The dataset is composed of twins--some of them are monozygotic and some of them are dizygotic.

Each subject has a disease or is healthy. Notably, some twin pairs are discordant (one is healthy and the other has disease) while some twin pairs are concordant (both have disease). Both mono and dizygotic pairs have examples that are either concordant and discordant.

The disease also has two subtypes--in some cases one twin has one disease subtype and the other twin has the other disease subtype.

The main comparison of interest is between healthy and disease, with secondary comparisons between heathy and each disease subtype, and between the two disease subtypes. Ideally I'd like to account for the pairs of twins in the design formula (maybe with blocking?). Will it work to include both concordant and discordant twins in the same analysis?

And considering that I suspect mono and dizygotic twins might have different factors differentiating healthy and disease, does it make sense to model the mono and dizygotic twins together? I'd be interested in seeing overall effects (associated with disease in both mono and dizygotic) and distinct effects (more strongly associated with disease in either mono or dizygotic twins).

Any guidance would be appreciated! The limma user manual is quite helpful, but I'm not sure where to go for a more complicated case like this one.