weirauchlab / reli Goto Github PK

The "Other platforms" section of the README needs step-by-step instructions like

1. download Eclipse CDT from link
2. rename CDT settings file to some other filename
3. open project file
4. click, this, that, and the other

Otherwise, we need to include a Makefile that works with Clang or gcc on macOS, and forget all about the Eclipse/Windows angle.

Under How to cite, you write that your paper is still under review but it's not. Right? :-)

Hi there,

I have a couple of questions about RELI. Firstly, I'm curious why LD is optional. Does this mean that RELI has a default LD file from 1000G? I tried running RELI without the -ld flag or leaving it empty, and none of the options works.

Secondly, in the paper's method section, it states that an observed intersection is recorded between each LD block and each dataset, and there would be a p-value for each dataset, right? However, doesn't the target label mean that we are only testing one specific dataset against null model? Would I need to use a for loop to iterate through each index?

Thank you for your time!

Best regards,
Sean

With the current release of RELI available on GitHub, you will get the following error messages if you attempt to make test on a freshly-cloned copy of the repository:

/bin/sh: 2: data/validate.sh: not found

ACK!

Data validation failed. Try deleting the 'data' directory
and running 'make test' again.

pushd example && ./example_run.sh
/bin/sh: 1: pushd: not found
Makefile:69: recipe for target 'test' failed
make: *** [test] Error 127

But stop! You DON'T need to delete the data directory! This error shows up because the packaging of the tarball on the server side has changed recently; the RELI repository will be updated shortly to reflect this change.

In the meantime, you can download the validation script manually, and re-run make test:

cd /path/to/cloned/repo
pushd data
curl -sLO https://tf.cchmc.org/external/RELI/validate.sh  # or use 'wget'
chmod a+x validate.sh
popd 
make test

You can ignore the /bin/sh: 1: pushd: not found error, if you receive one; that's just a benign bug that will be addressed in a future release.

I am trying to run RELI on a list of 3,635 genome-wide significant variants which are aggregated into 78 LD blocks, but after RELI loaded in the snp table, it failed when loading the LD table. The error I got is "Segmentation fault":

|---------------------------------------------------------------|
|                                                               |
|       Regulatory Element Locus Intersection (RELI) Analysis   |
|                       Current version: 0.90                   |
|                                                               |
|---------------------------------------------------------------|
Start Regulatory Element Locus Intersection (RELI) analysis.
Running arguements:
1) phenotype snp file: multiple_sclerosis.snp
2) phenotype LD structure file: multiple_sclerosis.ld
3) SNP matching mode: 0
4) null model file: ../data/Null/CommonSNP_MAFmatch
5) dbSNP table file: ../data/SNPtable/SNPtable
6) target chip-seq label: hg19_0302
7) chip-seq index file: ../data/ChIPseq.index
8) chip-seq data dir: ../data/ChIP-seq/
9) output dir name: Output/
10) genome build file:
11) statistics output file name: Output//hg19_0302.RELI.stats
12) overlapped locus numbers output file name: Output//hg19_0302.RELI.overlaps
13) overlapped snps output file name: Output//hg19_0302.RELI.rsids
14) provided phenotype name: MS
15) provided ancestry name: .
using default hg19 genome build
genome structure loaded.
chip-seq index file loaded.
target ChIP-seq file set.
target ChIP-seq file loaded, sorted, and width calculated.
null model loaded.
reading snp file completed.
snp table loaded.
snp MAF information queried.
phenotype LD file loaded.
Segmentation fault

I followed the format of SLE_EU.ld in the example/ folder but I am not sure if I am missing something. My .ld file contains 78 rows, each row starts with the top hit variant of the LD block, and is followed by ":" and a list of rs IDs in the same LD block with genome-wide significant pvalue. All rsIDs in the .ld file are listed in the .snp file (BED4 format).

I tried to run RELI without the .ld file (only run with .snp). It finished without any errors, but of course the results were not legitimate because LD was ignored - I got a Ratio of 1 (100% intersect) and p-value of 0.

Please advise on how to properly account for LD in the analysis. Thank you very much!

If we expect anyone to be able to do anything useful with the Docker image, the README needs to go into enough detail about bind-mounting the input and output directories, so that people can run it on their own data, and write the output to a directory outside the docker image.

Beyond that, the Docker image is only useful for running the one example from the CWL YAML input file, because CWL handles the bind mounts transparently to the user.

Hi I wonder if your group would consider releasing ChIP-seq and SNPtables in hg38 coordinates? Many recently available GWAS/ ChIP-seq data are of build 38, it would be great if RELI can support the analyses of those. Thanks!