A bug in ieugwasr::ld_matrix_local

[sskimb]

The following is a result of reading a bim file generated by plink, as implemented in ld_matrix_local.

  V1         V2 V3        V4   V5 V6
1  1  rs3811450  0 154551032 TRUE  C
2  1 rs11264222  0 154554357 TRUE  C
3  1     rs9616  0 154555733 TRUE  A

In this case, V5 has all "T" alleles. However, read.table in R guessed that this column must be a logical one as it is filled with only "T".
This behavior can be suppressed only in R 4.3.0+ by supplying an option "tryLogical=F", whose result is:

  V1         V2 V3        V4 V5 V6
1  1  rs3811450  0 154551032  T  C
2  1 rs11264222  0 154554357  T  C
3  1     rs9616  0 154555733  T  A

[laleoarrow]

Is this fixed yet?

[laleoarrow]

Im using R4.3.3, I did not experience this issue.

[sskimb]

Try to read a file, which has all "T" in the 5-th column such as: $ cat test.bim 1 rs38 0 32 T C 1 rs11 0 357 T C 1 rs96 0 733 T A When I try it, I get the followings: > bim <- read.table("test.bim", stringsAsFactors=F) > bim2 <- read.table("test.bim", stringsAsFactors=F, tryLogical=F) > bim == bim2 V1 V2 V3 V4 V5 V6 [1,] TRUE TRUE TRUE TRUE FALSE TRUE [2,] TRUE TRUE TRUE TRUE FALSE TRUE [3,] TRUE TRUE TRUE TRUE FALSE TRUE Notice that the 5-th column is all FALSE. This behavior does not affect the ld calculation by plink, as the bim is not used by plink in ieugwasr::ld_matrix_local. When the function returns the ld matrix, it mislabels the columns and rows when "with_alleles=TRUE". The behavior is not new, as it has been reported at https://www.biostars.org/p/258586/ The R-4.3.0 introduced "tryLogical", according to the release note: "type.convert() and hence read.table() get new option tryLogical = TRUE with back compatible default. When set to false, converts "F" or "T" columns to character." I am not in the position to update the github source code of ieugwasr. Instead, I made a workaround to post-process the ld matrix by fixing the column and row labels. I incorporated this fix into the TwoSampleMR::dat_to_MRInput function. If anybody is interested in this fix, please let me know. Sincerely, Sangsoo Kim 2024년 4월 15일 (월) 오전 10:53, leoarrow1 ***@***.***>님이 작성:

…

Im using R4.3.3, I did not experience this issue. image.png (view on web) <https://github.com/MRCIEU/ieugwasr/assets/127536391/a2050ee2-2aaf-43af-85a1-094a054035f7> — Reply to this email directly, view it on GitHub <#38 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADAJTYXURWG3OISIBW3BY6DY5MXJFAVCNFSM6AAAAAA4U5PQD2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANJUGMYDINJUHA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[laleoarrow]

Thank you for your prompt response. I have carefully reviewed your message and the provided code snippet. If Im getting the code right, to clarify, the ld_matrix_local function utilizes PLINK to extract genotype data for a given SNP (RS ID) from the reference panel. This below code is a snippet from the source for this process: ``` # source code for your convenience to check fun1 <- paste0(shQuote(plink_bin, type=shell), " --bfile ", shQuote(bfile, type=shell),   " --extract ", shQuote(fn, type=shell),   " --make-just-bim ",   " --keep-allele-order ",   " --out ", shQuote(fn, type=shell) ) system(fun1) ``` Subsequently, the BIM file is read using read.table:`bim <- read.table(paste0(fn, ".bim"), stringsAsFactors=FALSE)`, where bim file were used?  If `with_alleles = T`, it will also use him file to attach the allele name in rownames and colnames. Hence, the bim file is used in `ld_matrix_local`. I am currently encountering difficulties with the SuSiE-coloc analysis, as it indicates that the LD matrix is not consistent with the z-scores of the summary data. I am eager to learn more about your method for calculating the LD matrix, as I am still struggling with this aspect in order to perform a coloc-susie analysis. BTW the attached file and the image mentioned in the correspondence were not received. Thank you for the assistance. Best, Ao

[sskimb]

dat2MRInput.R.txt

This file is an R script, containing dat2MRInput and Harmonise_LD_dat functions.
In my procedure, both functions replace the corresponding functions, dat_to_MRInput and harmonise_ld_dat in the TwoSampleMR package.

Sincerely.

[laleoarrow]

dat2MRInput.R.txt

This file is an R script, containing dat2MRInput and Harmonise_LD_dat functions. In my procedure, both functions replace the corresponding functions, dat_to_MRInput and harmonise_ld_dat in the TwoSampleMR package.

Sincerely.

Really appreciate it! Thats really helpful!

Thanks,

Ao