Why are they different?

[yikeshu0611]

library(TwoSampleMR)
d1 <- extract_instruments('ieu-a-299', clump=T)

d2i <- extract_instruments('ieu-a-299',clump = F)
d2 <- clump_data(d2i)

d1$SNP[!d1$SNP %in% d2$SNP]
[1] "rs247616" "rs765548" "rs12286037" "rs2160669"
[5] "rs326214" "rs174535" "rs3829502" "rs289745"
[9] "rs1619021" "rs2972146" "rs2255141" "rs3996352"
[13] "rs12328675" "rs16842" "rs1279840" "rs10282707"
[17] "rs2240327" "rs634869" "rs9491696" "rs3936511"
[21] "rs2290547" "rs1942880"

[yikeshu0611]

force_server not work in extract_instruments
becasue

d <- ieugwasr::tophits(outcomes, pval = p1, clump = clump, 
    r2 = r2, kb = kb, force_server = FALSE, opengwas_jwt = opengwas_jwt, 
    x_api_source = x_api_source_header())

so why give force_server in the argument

[remlapmot]

Thanks for reporting this.

You were correct that the force_server argument was not passed correctly to the ieugwasr::tophits() call - that was an accidental bug. And in looking at the code I also found an issue with the clump argument.

So you can install the new version (0.6.26) from GitHub or after 1 hr from our mrcieu.r-universe.dev . However, in terms of your example I still get the same difference you see, so I'll have to ask one of the data experts about that.

[yikeshu0611]

Thanks for reporting this.

You were correct that the force_server argument was not passed correctly to the ieugwasr::tophits() call - that was an accidental bug. And in looking at the code I also found an issue with the clump argument.

So you can install the new version (0.6.26) from GitHub or after 1 hr from our mrcieu.r-universe.dev . However, in terms of your example I still get the same difference you see, so I'll have to ask one of the data experts about that.

Similarly, the p2 parameter in the extract_instruments function is not passed down.

[yikeshu0611]

In the mv_multiple function, the index has no effect by default, but it is passed to nsnp.

[remlapmot]

Thanks for reporting this.
You were correct that the force_server argument was not passed correctly to the ieugwasr::tophits() call - that was an accidental bug. And in looking at the code I also found an issue with the clump argument.
So you can install the new version (0.6.26) from GitHub or after 1 hr from our mrcieu.r-universe.dev . However, in terms of your example I still get the same difference you see, so I'll have to ask one of the data experts about that.

Similarly, the p2 parameter in the extract_instruments function is not passed down.

Thanks for this - yes that accidentally did nothing, as indeed did the clump_p2 argument to clump_data(); so I have removed both of those from the function definitions and docs.

You can install v0.6.27 with those removed.

[remlapmot]

In the mv_multiple function, the index has no effect by default, but it is passed to nsnp.

I accept there may well be bugs in some of the mv functions. On reading the code I couldn't immediately spot an issue with index there - but I could be wrong. Is there any chance you can provide an example which you think is wrong.

[remlapmot]

Thanks for reporting this.

You were correct that the force_server argument was not passed correctly to the ieugwasr::tophits() call - that was an accidental bug. And in looking at the code I also found an issue with the clump argument.

So you can install the new version (0.6.26) from GitHub or after 1 hr from our mrcieu.r-universe.dev . However, in terms of your example I still get the same difference you see, so I'll have to ask one of the data experts about that.

About the difference in the data - this could be due to differences in the LD reference panel used. The data folk are apparently working on some updates which would fix this.

[remlapmot]

library(TwoSampleMR) d1 <- extract_instruments('ieu-a-299', clump=T)

d2i <- extract_instruments('ieu-a-299',clump = F) d2 <- clump_data(d2i)

d1$SNP[!d1$SNP %in% d2$SNP] [1] "rs247616" "rs765548" "rs12286037" "rs2160669" [5] "rs326214" "rs174535" "rs3829502" "rs289745" [9] "rs1619021" "rs2972146" "rs2255141" "rs3996352" [13] "rs12328675" "rs16842" "rs1279840" "rs10282707" [17] "rs2240327" "rs634869" "rs9491696" "rs3936511" [21] "rs2290547" "rs1942880"

It turns out on the server for the clumping we started using 1/abs(z) as the exposure pvalue. In order to obtain reproducible results you need to amend the pval.exposure column as follows (admittedly there is still a one SNP difference).

library(TwoSampleMR)
#> TwoSampleMR version 0.6.29
#>   [>] New authentication requirements: https://mrcieu.github.io/ieugwasr/articles/guide.html#authentication.
#>   [>] Major upgrades to our servers completed to improve service and stability.
#>   [>] We need your help to shape our emerging roadmap!
#>       Please take 2 minutes to give us feedback -
#>       https://forms.office.com/e/eSr7EFAfCG
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
d1 <- extract_instruments('ieu-a-299', clump = 1)

d2i <- extract_instruments('ieu-a-299', clump = 0)
d2i <- d2i %>% mutate(pval.exposure = se.exposure / abs(beta.exposure))
d2 <- clump_data(d2i)
#> Please look at vignettes for options on running this locally if you need to run many instances of this command.
#> Clumping ieu-a-299, 3481 variants, using EUR population reference
#> Removing 3394 of 3481 variants due to LD with other variants or absence from LD reference panel

d1$SNP[!(d1$SNP %in% d2$SNP)]
#> [1] "rs2290547"
d2$SNP[!(d2$SNP %in% d1$SNP)]
#> [1] "rs12801636"

^{Created on 2025-12-19 with reprex v2.1.1}

Apologies that this is not documented. We will document this in the new year.

[JasonQiu]

library(TwoSampleMR)
library(dplyr)

# Reproducing the issue

dtophits <- extract_instruments('ieu-a-299')
dfull <- extract_instruments('ieu-a-299', clump=0)
dfull_z <- dfull %>% mutate(pval.exposure = se.exposure / abs(beta.exposure))
dclumpnew_z <- clump_data(dfull_z)

length(intersect(dtophits$SNP, dclumpnew_z$SNP))
# [1] 86
setdiff(dtophits$SNP, dclumpnew_z$SNP)
# [1] "rs2290547"
setdiff(dclumpnew_z$SNP, dtophits$SNP)
# [1] "rs12801636"

# Specify the legacy population

dclumpnew_z_legacy <- clump_data(dfull_z, pop="legacy")

identical(dtophits$SNP, dclumpnew_z_legacy$SNP)
# [1] TRUE

A “legacy” bfile (data_maf0.01_rs) has been used by the old and new pipelines, which means the tophits we have are based on this legacy bfile. However the default pop is now EUR in all packages and API e.g. as explained in TwoSampleMR. You should be able to reproduce our tophits if pop="legacy" is specified.

We are planning to rerun the tophits using the EUR population as part of the b38 migration process, and you may still see a tiny discrepancy on the results as above until then.