LD matrix

Hello,

I'm using the coloc susie function to study colocalisation of two disorders. I have calculated LD-matrices of 1000Genomes data with PLINK for both disorders. The number of SNPs in the datasets matches the number of SNPs in the LD matrices. I've added these matrices to my datasets using:

minimum_data1 <- as.list(dataset[c("beta","varbeta","p","SNP","POS")])
minimum_data1$type <- "cc"
minimum_data1$LD <- as.matrix(LD_data)

The data structure of my datasets looks just like the structure of the testdata.
Testdata:

str(D1$LD)
num [1:50, 1:50] 1 0.732 0.805 0.763 0.541 ...

attr(*, "dimnames")=List of 2
..$ : chr [1:50] "s1" "s2" "s3" "s4" ...
..$ : chr [1:50] "s1" "s2" "s3" "s4" ...
My data:

str(minimum_data1$LD)
num [1:3397, 1:3397] 1 1 1 0.0895 0.0804 ...

attr(*, "dimnames")=List of 2
..$ : chr [1:3397] "1:115581654" "1:115581717" "1:115581758" "1:115581922" ...
..$ : chr [1:3397] "1:115581654" "1:115581717" "1:115581758" "1:115581922" ...

When I use the command:

check_dataset(minimum_data1, suffix = "", req = c("SNP", "LD"), warn.minp = 1e-06)
It returns:
NULL

However, when I try to run runsusie, I get the error: Error in (function (z, R, z_ld_weight = 0, L = 10, prior_variance = 50, : The dimension of correlation matrix (0 by 0) does not agree with expected (3397 by 3397). The matrix however in fact is a matrix of 3397x3397. What am I doing wrong?

Thank you for your time,
Sophie

I'm not sure. Do the dimnames on your LD matrix match your SNP names exactly?

…

On Tue, 2021-10-12 at 03:05 -0700, riesmeijersa wrote: Hello, I'm using the coloc susie function to study colocalisation of two disorders. I have calculated LD-matrices of 1000Genomes data with PLINK for both disorders. The number of SNPs in the datasets matches the number of SNPs in the LD matrices. I've added these matrices to my datasets using: > minimum_data1 <- > as.list(dataset[c("beta","varbeta","p","SNP","POS")]) > minimum_data1$type <- "cc" > minimum_data1$LD <- as.matrix(LD_data) The data structure of my datasets looks just like the structure of the testdata. Testdata: > str(D1$LD) > num [1:50, 1:50] 1 0.732 0.805 0.763 0.541 ... * attr(*, "dimnames")=List of 2 ..$ : chr [1:50] "s1" "s2" "s3" "s4" ... ..$ : chr [1:50] "s1" "s2" "s3" "s4" ... My data: > str(minimum_data1$LD) > num [1:3397, 1:3397] 1 1 1 0.0895 0.0804 ... * attr(*, "dimnames")=List of 2 ..$ : chr [1:3397] "1:115581654" "1:115581717" "1:115581758" "1:115581922" ... ..$ : chr [1:3397] "1:115581654" "1:115581717" "1:115581758" "1:115581922" ... When I use the command: > check_dataset(minimum_data1, suffix = "", req = c("SNP", "LD"), > warn.minp = 1e-06) > It returns: > NULL However, when I try to run runsusie, I get the error: Error in (function (z, R, z_ld_weight = 0, L = 10, prior_variance = 50, : The dimension of correlation matrix (0 by 0) does not agree with expected (3397 by 3397). The matrix however in fact is a matrix of 3397x3397. What am I doing wrong? Thank you for your time, Sophie — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

Yes, the LD column and row names match the SNPs from the dataset exactly and both are ordered similarly:

str(minimum_data1)
List of 7
$ beta : num [1:3397] -0.0067 -0.0066 0.0067 0.002 0.0093 -0.0216 -0.003 0.0182 -0.0005 -0.0213 ...
$ varbeta: num [1:3397] 0.001884 0.001884 0.001884 0.000313 0.000335 ...
$ p : num [1:3397] 0.877 0.879 0.877 0.91 0.611 ...
$ SNP : chr [1:3397] "1:115581654" "1:115581717" "1:115581758" "1:115581922" ...
...
str(minimum_data1$LD)
num [1:3397, 1:3397] 1 1 1 0.0895 0.0804 ...

attr(*, "dimnames")=List of 2
..$ : chr [1:3397] "1:115581654" "1:115581717" "1:115581758" "1:115581922" ...
..$ : chr [1:3397] "1:115581654" "1:115581717" "1:115581758" "1:115581922" ...

then I am at a loss, unless you can share your data?

…

On Tue, 2021-10-12 at 04:42 -0700, riesmeijersa wrote: Yes, the LD column and row names match the SNPs from the dataset exactly and both are ordered similarly: > str(minimum_data1) > List of 7 > $ beta : num [1:3397] -0.0067 -0.0066 0.0067 0.002 0.0093 -0.0216 - > 0.003 0.0182 -0.0005 -0.0213 ... > $ varbeta: num [1:3397] 0.001884 0.001884 0.001884 0.000313 > 0.000335 ... > $ p : num [1:3397] 0.877 0.879 0.877 0.91 0.611 ... > $ SNP : chr [1:3397] "1:115581654" "1:115581717" "1:115581758" > "1:115581922" ... > ... > str(minimum_data1$LD) > num [1:3397, 1:3397] 1 1 1 0.0895 0.0804 ... * attr(*, "dimnames")=List of 2 ..$ : chr [1:3397] "1:115581654" "1:115581717" "1:115581758" "1:115581922" ... ..$ : chr [1:3397] "1:115581654" "1:115581717" "1:115581758" "1:115581922" ... — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

I can send them to you. What's your e-mail adress?

cew54 at cam.ac.uk

…

On Tue, 2021-10-12 at 05:10 -0700, riesmeijersa wrote: I can send them to you. What's your e-mail adress? — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

I suspect the issue is due to your input list naming the element "SNP" and not "snp", which causes the LD matrix to be subset to 0 rows and columns:

coloc/R/susie.R

Line 460 in f0c4a6a

LD=d$LD[d$snp,d$snp] # just in case

does that fix your issue @riesmeijersa ?

Yes it did

I ran into a similar issue. My PR #62 enables coloc to detect these data formatting issues early, and prevent passing an empty LD matrix tosusie_rss().

Thanks I just need time to test it before I pull https://chr1swallace.github.io

…

________________________________ From: John Blischak ***@***.***> Sent: Friday, November 19, 2021 8:53:11 PM To: chr1swallace/coloc ***@***.***> Cc: Chris Wallace ***@***.***>; Comment ***@***.***> Subject: Re: [chr1swallace/coloc] LD matrix (#59) I ran into a similar issue. My PR #62<#62> enables coloc to detect these data formatting issues early, and prevent passing an empty LD matrix tosusie_rss(). — You are receiving this because you commented. Reply to this email directly, view it on GitHub<#59 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AAQWR2EN53GKO74RBZLD6E3UM22LPANCNFSM5F2FTSIA>. Triage notifications on the go with GitHub Mobile for iOS<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

Two other, somewhat related questions:

@riesmeijersa: which command did you use in plink for this?
@chr1swallace: do we need to have r or r2?

For susie you should use --r for plink:

stephenslab/susieR#135 (comment)

thanks for replying @mhaiyue . yes, we need the correlation matrix, so r. Values should range between -1 and +1.