LDpred2 imputation

[deepchocolate]

Fixes one part of #130. Closing #138

Changes proposed in this pull request:

• Added an imputation script usecases/LDpred2/imputeGenotypes.R. Using bigsnpr::fastImputeSimple one can choose one of these methods, but I've also added zero "imputation" as described in the docs for this function.
• Changed the allele flipping in ldpred2.R and calculateLD.R as described in LDpred2 issue with MoBa data(?) #124
• Removed --geno-imputefrom ldpred2.R. --geno-impute-zero can be passed instead.
• Moved functionality into a separate script usecases/LDpred2/fun.R to enable reuse and improved testing LDpred2_example/unittests/fun.R.

Before submitting

• I've read and followed all steps in the Making a pull request
  section of the CONTRIBUTING docs.
• I've updated or added any relevant docstrings following the syntax described in the
  Writing docstrings section of the CONTRIBUTING docs.
• If this PR fixes a bug, I've added a test that will fail without my fix.
• If this PR adds a new feature, I've added tests that sufficiently cover my new functionality.

[deepchocolate]

@ofrei @espenhgn I tried adding fastImpute (in addition to fastImputeSimple), but this depends on xgboost (there's an R-package for it). However, I only get errors when trying to use it. Shall I stick with the simple?

[espenhgn]

@ofrei @espenhgn I tried adding fastImpute (in addition to fastImputeSimple), but this depends on xgboost (there's an R-package for it). However, I only get errors when trying to use it. Shall I stick with the simple?

I'm sure we can include xgboost in the r.sif build if this fixes the issue. Perhaps a specific version of this package is required for fastImpute?

[deepchocolate]

@ofrei @espenhgn I tried adding fastImpute (in addition to fastImputeSimple), but this depends on xgboost (there's an R-package for it). However, I only get errors when trying to use it. Shall I stick with the simple?

I'm sure we can include xgboost in the r.sif build if this fixes the issue. Perhaps a specific version of this package is required for fastImpute?

I'll stick with fastImputeSimple for now as there are some other isues with the scripts, that needs a fix.

[espenhgn]

Thanks a lot! I encountered some minor issues which were easily resolved (locally), as commented. You can merge #139 into this PR. Other than that:

• in the LDpred2/README.md height example, add the line $RSCRIPT imputeGenotypes.R --impute-simple mean0 --geno-file-rds $fileGenoRDS after creating $fileGenoRDS itself. Otherwise it won't run. Either this, or add the --geno-impute-zero argument to $RSCRIPT ldpred2.R ... calls below
• It was my bad suggestion to call the test directory for LDpred2_example. Perhaps this can by renamed LDpred2_test in this PR?

[espenhgn]

Not sure if it's a concern, but for this entire call a warning is thrown while merging dataframes, possibly on L220-222 in the LDpred2.R script:

Test no restrictions on snps (similar to tutorial)
Loading backingfile: /nrec/space/espenh/containers/usecases/LDpred2_example/data/public-data3.rds 

### Reading LD reference meta-file from  /nrec/space/espenh/containers/usecases/LDpred2_example/output/ld/map.rds 

### Reading summary statistics /nrec/space/espenh/containers/usecases/LDpred2_tutorial/tutorial_data/public-data3-sumstats.txt 
Loaded 50000 SNPs
Filtering SNPs based on --chr2use
Retained 50000 out of 50000 
Matching sumstats to genotypes
50,000 variants to be matched.
0 ambiguous SNPs have been removed.
45,337 variants have been matched; 22,758 were flipped and 15,092 were reversed.
Matching sumstats to LD reference
45,337 variants to be matched.
0 ambiguous SNPs have been removed.
45,337 variants have been matched; 0 were flipped and 0 were reversed.
Warning message:
In merge.data.table(as.data.table(sumstats4), as.data.table(info_snp),  :
  column names 'pos.ss' are duplicated in the result
...

[deepchocolate]

@espenhgn I think I've addressed all points. I renamed usecases/LDpred2_example to usecases/LDpred2_test as you suggested

[espenhgn]

Pushed a small unzip fix to /LDpred2_test/run.sh. Looks good to me!

[deepchocolate]

Pushed a small unzip fix to /LDpred2_test/run.sh. Looks good to me!

Great, thanks!

[deepchocolate]

@ofrei Are you ok with merging this?

[deepchocolate]

@espenhgn Shall I merge this or wait for an approval from @ofrei aswell?

[espenhgn]

@espenhgn Shall I merge this or wait for an approval from @ofrei aswell?

No need to wait for @ofrei. Just go ahead and squash-merge this one :)

[espenhgn]

Btw., it's fine if you merge this first. I expect a small conflict with #141 that I can fix before that one is merged.