mtag
mtag copied to clipboard
JonJala
use my own ref panel instead of the default
I am trying to use a difference reference panel and MTAG is not accepting the panel I currently have. the panel I have was made using the generic ldsc.py, could you direct me to either which file MTAG accept in terms of ref panel or how I can use MTAG built in ldsc_mod to compute my own? Thank you
When you say it wasn't accepting the panel you generated, what do you mean?
On Thu, Sep 28, 2023, 5:49 PM AudeIkuzwe @.***> wrote:
I am trying to use a difference reference panel and MTAG is not accepting the panel I currently have. the panel I have was made using the generic ldsc.py, could you direct me to either which file MTAG accept in terms of ref panel or how I can use MTAG built in ldsc_mod to compute my own? Thank you
— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/189, or unsubscribe https://github.com/notifications/unsubscribe-auth/APIOF52LVIBGDDZR2YI5CDTX4XWIDANCNFSM6AAAAAA5LTDIOI . You are receiving this because you are subscribed to this thread.Message ID: @.***>
this is the error I get when I use my own panel '# terms in --M must match # of LD Scores in --ref-ld.\n' + str(e.args)) ValueError: # terms in --M must match # of LD Scores in --ref-ld.
Would you mind sharing the contents of the log file? Or minimally, the command you used to run MTAG? Do you have a log for when you created your own panel, as well? It's hard to know what is mismatching, right now there's not a ton to go on.
On Fri, Sep 29, 2023 at 12:59 PM AudeIkuzwe @.***> wrote:
this is the error I get when I use my own panel '# terms in --M must match
of LD Scores in --ref-ld.\n' + str(e.args))
ValueError: # terms in --M must match # of LD Scores in --ref-ld.
— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/189#issuecomment-1741214472, or unsubscribe https://github.com/notifications/unsubscribe-auth/APIOF5Y447SKMMZFQ3S752TX4347PANCNFSM6AAAAAA5LTDIOI . You are receiving this because you commented.Message ID: @.***>
hi yes, here is the log file from that MTAG run, with the command I ran included tutorial_results_sep28.log
here is the command that created the reference panel I am feeding into the MTAG (note I changed the .parquet files into just .gz files before running MTAG) path/python path/compute_ldscores.py --bfile pathtoplobkfiles/subset.10 --ld-wind-cm 1 --out subset.10.l2.ldscore.parquet
here is the log file for the above command Applying SNPs filter... [INFO] After filtering, 45258 SNPs remain [INFO] Applying initial ldscores loop [INFO] Applying main ldscores loop...
Please let me know if you would like me to add more information
Thank you. Hmm...I don't see anything in what you're passing in that looks suspect. Not having written the code myself, I'm not entirely sure what is going wrong, but my understanding is that using generic ldsc.py should be totally fine and that ldsc_mod is mostly just different in ways that are performance-related.
hi again, I have tried using ldsc and ldsc mod and still no luck, would you know if there is any publicly available panel for African ancestry that one can use? Thank you.
Hi [AudeIkuzwe] Did you ever figure this out ? My input GWAS are all on African populations, so I was wondering if I should use a different reference panel as input to Mtag.
I believe the Price Lab has LD scores publicly available, but I'm not totally sure. We intend to post some ourselves at some point too, but I don't anticipate it will be very soon.
Sorry I can't be more helpful.
On Fri, Oct 27, 2023, 4:38 PM samreenzafer @.***> wrote:
Hi [AudeIkuzwe] Did you ever figure this out ? My input GWAS are all on African populations, so I was wondering if I should use a different reference panel as input to Mtag.
— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/189#issuecomment-1783485840, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBUB5OY6JCI2TNKP7ZSSLDYBQLTZAVCNFSM6AAAAAA5LTDIOKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTGQ4DKOBUGA . You are receiving this because you are subscribed to this thread.Message ID: @.***>
It did not work, I had to get the reference panel from a collaborator that was the only way it worked
I created my own LDscores for Afr ethnicity, by extracting reported Africans from the 1000G cohort, excluding a few related samples (detected by KING), and extracting the HapMap3 snps (using same one provided by LDSC).
After running MTAG with both Eur based LDscores (provided by mtag), and my own Afr based LDscores, I don't see much of a difference in the Omega and Sigma correlations matrix, nor much difference in the Mtag adjusted summary stats.
Not sure if this is an acceptable outcome. @AudeIkuzwe Did you try any such comparison analyses ?
On Wed, Nov 1, 2023 at 7:26 AM AudeIkuzwe @.***> wrote:
It did not work, I had to get the reference panel from a collaborator that was the only way it worked
— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/189#issuecomment-1789053631, or unsubscribe https://github.com/notifications/unsubscribe-auth/AADLD63U2BFIJXCM5Q65T4LYCJL2XAVCNFSM6AAAAAA5LTDIOKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBZGA2TGNRTGE . You are receiving this because you commented.Message ID: @.***>
-- Regards,
Samreen
Thanks for replying, none are provided by the Price lab, so I created my own , please see my previous comment to @AudeIkuzwe
I'm not sure If my method and results are correct. If you have a suggestion or correction to my method, please let me know. Each of my Gwas comes from the same cohorts (hence sample overlap is high, especially most of the controls). Power of each GWAs analysis is not so high.
Trait1: 1172 are cases and 3337 are controls Trait2: 231 are cases and 3337 are controls. Trait3: 933 are cases and 3284 are controls Trait4: 964 are cases and 3253 are controls
Summary of MTAG results:
Trait # SNPs used N (max) N (mean) GWAS mean chi^2 MTAG mean chi^2 GWAS equiv. (max) N
1 ....ADD.summary.txt 11527276 4509 4509 1.001 1.006 25980
2 ...dHWE.summary.txt 11527276 3568 3568 0.996 1.000 -279
3 ...stic.summary.txt 11527276 4217 4217 0.993 0.995 3447
4 ...stic.summary.txt 11527276 4217 4217 0.996 0.994 5667
Estimated Omega: [[ 2.416e-07 -3.839e-11 -5.021e-11 3.227e-11] [-3.839e-11 -1.018e-06 -5.454e-11 -3.161e-11] [-5.021e-11 -5.454e-11 -1.599e-06 1.317e-11] [ 3.227e-11 -3.161e-11 1.317e-11 -9.868e-07]] (Correlation): [[ 1. nan nan nan] [nan nan nan nan] [nan nan nan nan] [nan nan nan nan]] Estimated Sigma: [[1.005 0.498 0.218 0.165] [0.498 0.995 0.11 0.125] [0.218 0.11 1.003 0.206] [0.165 0.125 0.206 1.016]] (Correlation): [[1. 0.498 0.218 0.163] [0.498 1. 0.11 0.125] [0.218 0.11 1. 0.204] [0.163 0.125 0.204 1. ]] MTAG weight factors: (average across SNPs) [0.358 0.408 0.635 0.668]
On Tue, Oct 31, 2023 at 4:57 PM paturley @.***> wrote:
I believe the Price Lab has LD scores publicly available, but I'm not totally sure. We intend to post some ourselves at some point too, but I don't anticipate it will be very soon.
Sorry I can't be more helpful.
On Fri, Oct 27, 2023, 4:38 PM samreenzafer @.***> wrote:
Hi [AudeIkuzwe] Did you ever figure this out ? My input GWAS are all on African populations, so I was wondering if I should use a different reference panel as input to Mtag.
— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/189#issuecomment-1783485840, or unsubscribe < https://github.com/notifications/unsubscribe-auth/AFBUB5OY6JCI2TNKP7ZSSLDYBQLTZAVCNFSM6AAAAAA5LTDIOKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTGQ4DKOBUGA>
. You are receiving this because you are subscribed to this thread.Message ID: @.***>
— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/189#issuecomment-1788192095, or unsubscribe https://github.com/notifications/unsubscribe-auth/AADLD653OTTPXNMXMHCX2LLYCGF7FAVCNFSM6AAAAAA5LTDIOKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBYGE4TEMBZGU . You are receiving this because you commented.Message ID: @.***>
-- Regards,
Samreen