Missing SNPs in MTAG summary results

[JuRaGa]

Dear Omeed,

I have two phenotypes that are moderately correlated (rho = 0.6). Therefore, I used MTAG to boost the power of my single-GWASs. However, when inspecting the results, I see that some of the original SNPs available in the single-GWAS summary statistics are missing from the MTAG results for both phenotypes.

I have checked for allele inconsistency between both single-GWASs and this is not the case.

I was wondering if you could help me with this issue?

Many thanks, Julia

[paturley]

Hi Julia,

There are many filters that MTAG imposes beyond allelic consistency. It's hard to diagnose the problem without seeing the data. Can you share the row of the summary statistic file corresponding to a SNP that is present in both GWAS files and not present in the MTAG file?

Best, Patrick

On Tue, Dec 18, 2018 at 11:07 AM Julia-Ramirez [email protected] wrote:

Dear Omeed,

I have two phenotypes that are moderately correlated (rho = 0.6). Therefore, I used MTAG to boost the power of my single-GWASs. However, when inspecting the results, I see that some of the original SNPs available in the single-GWAS summary statistics are missing from the MTAG results for both phenotypes.

I have checked for allele inconsistency between both single-GWASs and this is not the case.

I was wondering if you could help me with this issue?

Many thanks, Julia

— 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/59, or mute the thread https://github.com/notifications/unsubscribe-auth/AUNA9UNWjVk-dWJ1aRku96JoGhHa4o3uks5u6RKwgaJpZM4ZYmoF .

[JuRaGa]

Hi Patrick,

Thanks for your quick reply.

This is the row from the single-GWAS summary statistics for phenotype 1:

         snpid            CHR        BP       a1 a2 A1FREQ     pval           n                z

13:74505951_CT_C 13 74505951 CT C 0.365313 9.6e-14 49711.36 7.446459

This is the row from the single-GWAS summary statistics for phenotype 2: snpid CHR BP a1 a2 A1FREQ pval n z

13:74505951_CT_C 13 74505951 CT C 0.365748 0.012 49346.31 2.51686390581907

And then this SNP is missing from the MTAG summary statistics for both phenotypes.

I also have examples for SNVs that are not indels, so this doesn't only happen with indels.

Thanks very much for your help,

Julia

On 18/12/2018 21:39, paturley wrote: Hi Julia,

There are many filters that MTAG imposes beyond allelic consistency. It's hard to diagnose the problem without seeing the data. Can you share the row of the summary statistic file corresponding to a SNP that is present in both GWAS files and not present in the MTAG file?

Best, Patrick

On Tue, Dec 18, 2018 at 11:07 AM Julia-Ramirez [email protected]mailto:[email protected] wrote:

Dear Omeed,

I have two phenotypes that are moderately correlated (rho = 0.6). Therefore, I used MTAG to boost the power of my single-GWASs. However, when inspecting the results, I see that some of the original SNPs available in the single-GWAS summary statistics are missing from the MTAG results for both phenotypes.

I have checked for allele inconsistency between both single-GWASs and this is not the case.

I was wondering if you could help me with this issue?

Many thanks, Julia

— 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/59https://github.com/omeed-maghzian/mtag/issues/59, or mute the thread https://github.com/notifications/unsubscribe-auth/AUNA9UNWjVk-dWJ1aRku96JoGhHa4o3uks5u6RKwgaJpZM4ZYmoFhttps://github.com/notifications/unsubscribe-auth/AUNA9UNWjVk-dWJ1aRku96JoGhHa4o3uks5u6RKwgaJpZM4ZYmoF .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/omeed-maghzian/mtag/issues/59#issuecomment-448380406, or mute the threadhttps://github.com/notifications/unsubscribe-auth/Aij-i3BmU8G-zy5Ny3AqG89saG2VDj-Uks5u6WCqgaJpZM4ZYmoF.

[https://ipmcdn.avast.com/images/icons/icon-envelope-tick-round-orange-animated-no-repeat-v1.gif]https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient Virus-free. www.avast.comhttps://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient

[paturley]

MTAG includes a number of filters by default, including any filters imposed by ldsc. (Which includes indels, I believe.) Have you looked over the documentation to confirm that there are not filters in place that cause you to lose those SNPs?

On Wed, Dec 19, 2018, 3:28 AM Julia-Ramirez <[email protected] wrote:

Hi Patrick,

Thanks for your quick reply.

This is the row from the single-GWAS summary statistics for phenotype 1:

snpid CHR BP a1 a2 A1FREQ pval n z

13:74505951_CT_C 13 74505951 CT C 0.365313 9.6e-14 49711.36 7.446459

This is the row from the single-GWAS summary statistics for phenotype 2: snpid CHR BP a1 a2 A1FREQ pval n z

13:74505951_CT_C 13 74505951 CT C 0.365748 0.012 49346.31 2.51686390581907

And then this SNP is missing from the MTAG summary statistics for both phenotypes.

I also have examples for SNVs that are not indels, so this doesn't only happen with indels.

Thanks very much for your help,

Julia

On 18/12/2018 21:39, paturley wrote: Hi Julia,

There are many filters that MTAG imposes beyond allelic consistency. It's hard to diagnose the problem without seeing the data. Can you share the row of the summary statistic file corresponding to a SNP that is present in both GWAS files and not present in the MTAG file?

Best, Patrick

On Tue, Dec 18, 2018 at 11:07 AM Julia-Ramirez <[email protected]

mailto:[email protected] wrote:

Dear Omeed,

I have two phenotypes that are moderately correlated (rho = 0.6). Therefore, I used MTAG to boost the power of my single-GWASs. However, when inspecting the results, I see that some of the original SNPs available in the single-GWAS summary statistics are missing from the MTAG results for both phenotypes.

I have checked for allele inconsistency between both single-GWASs and this is not the case.

I was wondering if you could help me with this issue?

Many thanks, Julia

— 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/59< https://github.com/omeed-maghzian/mtag/issues/59>, or mute the thread < https://github.com/notifications/unsubscribe-auth/AUNA9UNWjVk-dWJ1aRku96JoGhHa4o3uks5u6RKwgaJpZM4ZYmoF < https://github.com/notifications/unsubscribe-auth/AUNA9UNWjVk-dWJ1aRku96JoGhHa4o3uks5u6RKwgaJpZM4ZYmoF>

.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub< https://github.com/omeed-maghzian/mtag/issues/59#issuecomment-448380406>, or mute the thread< https://github.com/notifications/unsubscribe-auth/Aij-i3BmU8G-zy5Ny3AqG89saG2VDj-Uks5u6WCqgaJpZM4ZYmoF>.

[ https://ipmcdn.avast.com/images/icons/icon-envelope-tick-round-orange-animated-no-repeat-v1.gif ]< https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient> Virus-free. www.avast.com< https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient>

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/omeed-maghzian/mtag/issues/59#issuecomment-448510089, or mute the thread https://github.com/notifications/unsubscribe-auth/AUNA9VwFVKSmqro2ux5f3A0iujwiNjnYks5u6fiogaJpZM4ZYmoF .

[huilisabrina]

Hi @Julia-Ramirez ,

MTAG adopts some of the basic filters used in the munging step of ldsc, one of which is dropping multiallelic variants.

For the specific SNP you mentioned, it was most likely filtered at this step: (you can check this line in your log file. The [x] in your case is probably non-zero.

Removed [x] variants that were not SNPs. Note: strand ambiguous SNPs were not dropped.

However, if you really want to retain those multi-allelic variants, there is a way to do so. You can try modifying your local version of the MTAG code by turning on one of the (rarely-used) arguments it passes on to ldsc: You can change no_alleles=False to no_alleles=True in line [154] and [156] of mtag.py. I just tested it and this works for me.

Multiallelic variants can be retained in the MTAG output this way. For general purpose however, we won't change the current default option (i.e. filtering out multiallelic variants), but maybe we can add an option for the users in the future.

Let me know if this tweak works for you, or I can send you the modified copy via email.

Best, Hui

[JuRaGa]

Dear Hui,

Thank you very much for your reply. Does this multi-allelic filter apply to the effect allele only or also to the other allele? This could explain why these SNPs are filtered out.

Also, thank you very much for the advice with the flag. We will try this and let you know.

Merry Christmas and happy new year!

Julia

On 28/12/2018 23:37, huilisabrina wrote:

Hi @Julia-Ramirezhttps://github.com/Julia-Ramirez ,

MTAG adopts some of the basic filters used in the munging step of ldschttps://github.com/bulik/ldsc/wiki/Heritability-and-Genetic-Correlation#reformatting-summary-statistics, one of which is dropping multiallelic variants.

For the specific SNP you mentioned, it was most likely filtered at this step: (you can check this line in your log file. The [x] in your case is probably non-zero.

Removed [x] variants that were not SNPs. Note: strand ambiguous SNPs were not dropped.

However, if you really want to retain those multi-allelic variants, there is a way to do so. You can try modifying your local version of the MTAG code by turning on one of the (rarely-used) arguments it passes on to ldsc: You can change no_alleles=False to no_alleles=True in line [154] and [156] of mtag.py. I just tested it and this works for me.

Multiallelic variants can be retained in the MTAG output this way. For general purpose however, we won't change the current default option (i.e. filtering out multiallelic variants), but maybe we can add an option for the users in the future.

Let me know if this tweak works for you, or I can send you the modified copy via email.

Best, Hui

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/omeed-maghzian/mtag/issues/59#issuecomment-450443582, or mute the threadhttps://github.com/notifications/unsubscribe-auth/Aij-i7C8NvyKAtRJocjyk1sI2sVo04Vsks5u9qsxgaJpZM4ZYmoF.

[https://ipmcdn.avast.com/images/icons/icon-envelope-tick-round-orange-animated-no-repeat-v1.gif]https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient Virus-free. www.avast.comhttps://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient

[huilisabrina]

Hi @Julia-Ramirez ,

Yes the multi-allelic filter applies to both the effect and alternative allele. (@rkwalters please correct me if this/my understanding of the ldsc munging step is wrong!)

Happy new year!

Best, Hui