mtag
mtag copied to clipboard
JonJala
The traits selection
Dear sir, I have improved my previous mtag input after reading all the issues, I used the effect sample size to be the N column, but I still can't get more genome-wide significant SNPs than the original GWAS summary statistics. I wonder why I get worse results not just like the results in your paper. I guess maybe the rg of the traits 0.257 is to low. I calculated the rg of SWB and DEP,SWB and NEUR,-0.6757,-0.7656,respectively. The log file is here. AD_PA.log
If my idea is correct please let me know and give some advices about the genetic correlation of the triats I choose. I saw some isuues just like Memory Error with many traits #48 used many traits. I think the genetic correlation of the traits are not high, I used 13 random traits and also get less genome-wide significant SNP. I have been confused for a long time. If you can give me some advice, I will be very appreciate for that.
Best wishes, Crane
Hello Crane,
A key assumption of MTAG is that SNPs are exchangeable with respect to the correlation of effects sizes across traits. In practice, this is roughly equivalent to saying that the same set of SNPs are causal for each of the traits considered. If this is violated, MTAG may cause some loss of genome-wide significant SNPs because your are meta-analyzing SNPs that are truly associated with one trait and not associated with another. In the paper, we recommend that user not MTAG sets of traits together than have a genetic correlation of less than 0.7 unless your sample sizes are large enough.
On Mon, May 18, 2020 at 11:48 PM whecrane [email protected] wrote:
Dear sir, I have improved my previous mtag input after reading all the issues, I used the effect sample size to be the N column, but I still can't get more genome-wide significant SNPs than the original GWAS summary statistics. I wonder why I get worse results not just like the results in your paper. I guess maybe the rg of the traits 0.257 is to low. I calculated the rg of SWB and DEP,SWB and NEUR,-0.6757,-0.7656,respectively. If my idea is correct please let me know and give some advices about the genetic correlation of the triats I choose. I saw some isuues just like Memory Error with many traits #48 https://github.com/omeed-maghzian/mtag/issues/48 used many traits. I think the genetic correlation of the traits are not high, I used 13 random traits and also get less genome-wide significant SNP. I have been confused for a long time. If you can give me some advice, I will be very appreciate for that.
Best wishes, Crane
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/omeed-maghzian/mtag/issues/94, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBUB5NHSO5Y4RM7WEQC7OTRSH6RXANCNFSM4NEUB6BA .
Hi Paturley, Thank you very much for your reply, it helped me a lot. Have a nice day! Best wishes, Crane
Dear Paturley,
I downloaded some GWAS summary statistics, which have Neff but not have the column of freq, can I use the formula Neff=1/( 2p(1-p)(SE^2) ),(p is the MAF) to get MAF? I used the formula of this
p=(1+sqrt(1-2/(Neff(SE^2))))/2 or p=(1-sqrt(1-2/(Neff*(SE^2))))/2. Because there are two results for p, one is over 0.5, the other one is less than 0.5, I don't know how to solve it.
You will be so nice if you give me some advice for the p. Thank you very much. Have a nice day!
Best wishes, Crane
Hello Crane,
I would just use the MAF from a reference panel. Assuming you choose a panel with approximately the same ancestry, the allele frequencies should be the same.
On Tue, May 19, 2020 at 9:29 PM whecrane [email protected] wrote:
Dear Paturley,
I downloaded some GWAS summary statistics, which have Neff but not have the column of freq, can I use the formula Neff=1/( 2p(1-p) (SE^2) ),(p is the MAF) to get MAF? I used the formula of this p=(1+sqrt(1-2/(Neff(SE^2))))/2 or p=(1-sqrt(1-2/(Neff*(SE^2))))/2. Because there are two results for p, one is over 0.5, the other one is less than 0.5, I don't know how to solve it. You will be so nice if you give me some advice for the p. Thank you very much. Have a nice day!
Best wishes, Crane
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/omeed-maghzian/mtag/issues/94#issuecomment-631179452, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBUB5JRNBZLPGRRYTWQQ4DRSMW77ANCNFSM4NEUB6BA .
Hi Paturley, Thank you very much. I will follow your advice and have a try.
Best wishes, Crane