Following a two-factor GSEM with a commonfactor perSNP analyses for only 1 of the factors

Dear all,

I ran LDSC and a 2-factor Genomic SEM model including lymphoma subtypes and autoimmune diseases.

The LDSC results show limited genetic correlation between lymphomas and autoimmune diseases, but stronger correlations within each disease group. The 2-factor Genomic SEM model similarly suggests two largely separate latent factors, with a very small factor correlation, around -0.04.

I would now like to run per-SNP analyses. Ideally, I would use userGWAS() for the full 2-factor model, but I have not been able to get this working.

Since my main interest is lymphoma, would it be defensible to instead run a per-SNP common factor GWAS using only the lymphoma traits (this works fine in my data)?

So the analysis would be:

  1. LDSC across lymphoma and autoimmune diseases

  2. 2-factor Genomic SEM showing minimal correlation between factors

  3. Lymphoma-only common factor GWAS as the per-SNP follow-up

Does this seem like a reasonable strategy, or would it be problematic to move from the full 2-factor model to a lymphoma-only per-SNP model?

I am having trouble decifering if this would be seen as problematic.

Thank you for a great course!

Kind regards,
Jojo