Interpreting results after using "--incld_ambig_snps"

[Merr22]

Hello,

I am interested in the command “--incld_ambig_snps”, when I do not use this command MTAG removes over 600,000 SNPs out of 4,000,000 SNPs. I have tried MTAG both ways (with and without the “--incld_ambig_snps” command). In my case, in the MTAG output I have 100 significant (p<5x10^8) SNPs that are strand ambiguous for one trait that are not reported without the command. Can I report significant strand ambiguous SNPs as significant findings the same way as I might report other sig SNPs?

Additionally, I want to obtain the odds ratio for these SNPs to inform the trait risk associated with having each SNP - considering that the omega and sigma values that go into the MTAG calculation of beta are not including strand ambiguous values & considering that SNP odds ratios are calculated from SNP beta values, is it appropriate to use the beta of individual strand ambiguous SNPs to calculate the odds ratio for these SNPs?

Additionally, any advice surrounding why strand ambiguous SNPs are excluded & any information to help me learn more about the limitations of strand ambiguous SNP results would be extremely helpful. My own search has not turned up much.

Thank you,
Merritt

[paturley]

That option was inherited from the LDSC program on which MTAG is based. The point of excluding those is that there is a risk that the major and minor alleles are flipped and you wouldn't be able to tell that from the data. If you are confident that those SNPs are correctly labeled, then using the strand-ambiguous SNPs should be fine. You can treat MTAG betas the same way as GWAS betas, so transforming those to and odds ratio should be fine. Best, Patrick

…

On Wed, Jul 10, 2024 at 9:53 AM Merr22 ***@***.***> wrote: Hello, I am interested in the command “--incld_ambig_snps”, when I do not use this command MTAG removes over 600,000 SNPs out of 4,000,000 SNPs. I have tried MTAG both ways (with and without the “--incld_ambig_snps” command). In my case, in the MTAG output I have 100 significant (p<5x10^8) SNPs that are strand ambiguous for one trait that are not reported without the command. Can I report significant strand ambiguous SNPs as significant findings the same way as I might report other sig SNPs? Additionally, I want to obtain the odds ratio for these SNPs to inform the trait risk associated with having each SNP - considering that the omega and sigma values that go into the MTAG calculation of beta are not including strand ambiguous values & considering that SNP odds ratios are calculated from SNP beta values, is it appropriate to use the beta of individual strand ambiguous SNPs to calculate the odds ratio for these SNPs? Additionally, any advice surrounding why strand ambiguous SNPs are excluded & any information to help me learn more about the limitations of strand ambiguous SNP results would be extremely helpful. My own search has not turned up much. Thank you, Merritt — Reply to this email directly, view it on GitHub <#215>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFBUB5OV6WV5DNP4COWGCTLZLU4HJAVCNFSM6AAAAABKU67EZ2VHI2DSMVQWIX3LMV43ASLTON2WKOZSGQYDAOBTG43TAMQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[Merr22]

Thank you for your informative answer, this was very helpful.

Appreciatively,
Merritt