joepickrell/fgwas

Functional genomics and genome-wide association studies

fgwas

fgwas is a command line tool for integrating functional genomic information into a genome-wide association study (GWAS). The basic setup is as follows: you have performed a GWAS or a meta-analysis of many GWAS, and have identified tens of loci that influence the disease or trait (our approach works best if there are at least ~20 independent loci with p-values less than 5e-8). We set out to address the following questions:

1. Are these associations enriched in particular types of regions of the genome (coding exons, DNAse hypersensitive sites, etc.)?
2. Can we use these enrichments (if they exist) to identify novel loci influencing the trait?

The underlying model is described in: Pickrell JK (2014) Joint analysis of functional genomic data and genome-wide association studies of 18 human traits. bioRxiv 10.1101/000752

The annotations used in this paper are available from GitHub.

##Installation##

###Dependencies### fgwas depends on:

• the GNU Scientific Library

• the Boost Libraries

###Quick Start### The most up-to-date release is: version 0.3.6. See "Releases" above. After downloading fgwas-0.3.3.tar.gz at the link above, run:

tar -xvf fgwas-0.3.6.tar.gz

cd fgwas-0.3.6

./configure

make

This will create an executable file called fgwas in the src directory. The most common compilation error is that the configure script cannot find Boost or GSL. You may have to tell the script explicitly where to find them. For example, on OS X using macports, installations go to the non-standard path /opt/local/lib. To configure in this case, replace the above configure step with:

./configure LDFLAGS=-L/opt/local/lib

Example data is available in the test_data/ directory. To ensure that fgwas is working, run:

./src/fgwas -i test_data/test_LDL.fgwas_in.gz -w ens_coding_exon

A user guide is available here: fgwas v0.3.x User Guide

Previous versions are available from the Google Code repository.

fgwas's Issues

current segment is 70755198 73174848, position is 7

Hello,

Thanks for your great tool! I am using fgwas to calculate null correlatoins for gwas-pw and I ran into following error: "current segment is 70755198 73174848, position is 7"

It seems errors related to bed file, but I couldn't pinpoint the source of this error. I made sure all the SNPs were within the rages of segments (in the bed file) and ordered chromosome number and position appropriately. In addition, fgwas ran well for all the summary gwas stats I have other than this particular file.

Could somebody help me solve this issue?

Thank you very much in advance!

Details for using fGWAS

Dear Dr. Pickrell,
I'm confused about how to use fGWAS although I have read User Manual for fgwas v.0.3.x.
Would you mind giving me specific details or codes about incorporating GWAS summary datasets with 1000-genome?
Thank you.

Sincerely,
Wenqiang.

Annotation enrichments from fgwas - what is the base of the logartihm?

Hello,

In the .params output file, fgwas reports the enrichment values for each annotation used as well as the 95% confidence intervals. Are these enrichments on the log scale? If yes, what is the base of the logarithm used? Maybe I missed something, but I could not find it in the documentation

Kind regards,
Kaur

Input bayes factors

Could be useful to read in Bayes factors directly from the input file (e.g. from gwas software that give this as an output) rather than approximating using the Z-score/SE.

Interpretation of log likelihood

Dear Dr Pickrell,

I am currently using the FGWAS software which you developed, and have some questions:

1. Could you tell me how the ln(llk) value in the *.llk file is interpreted?
   a. For example we observed a ln(llk) value of 504 –does this correspond to a likelihood value of 504 to the base e or -504 to the base e?
2. In an analysis we are observing a ln(llk) value of greater than 43619.5. Is this a symptom of convergence issues: The dataset we are using does not have many strong signals.

Many Thanks,

Chris Grace

Bioinformatician
PROCARDIS Group
The Wellcome Trust Centre for Human Genetics
Roosevelt Drive, Oxford, OX3 7BN, United Kingdom

Average over SNP priors

Right now the Bayes factor for SNPs is calculate using a single prior on the effect size. In some cases it would be nice to average over BFs calculated using different priors.

NCASE and NCONTROL are not available in the summary statistics of a case-control study

Hi,

I would like to use this awesome package, however, I just realized NCASE and NCONTROL values were not included in the summary statistics of the case-control study. Could you please advise me on how to proceed?

Is there a way to infer these two values from other values? In the source code, it seems these two values are used to approximate V using approx_v_cc function, while there is also a commented line, V = SE * SE.
Are these two values not necessary if SE is known?
Or is it okay just pretend the case-control study is a quantitative one? Thanks a lot.

Convergence failures when there are a small number of significant hits in an annotation

If there are only a small number of GWAS hits in an annotation, fgwas sometimes fails to converge, as it "tries" to estimate the MLE of the enrichment parameter as -Inf. A simple fix is to only use the penalized likelihood (-onlyp -print).

A more involved approach would be to implement an explicit prior on the enrichment parameters and integrate over those.

Dealing with polymorphism

Hello,

It seems impossible to use fgwas on polyallelic sites.
I've got the following error message :
ERROR: SNPs out of order
Chromosome 1. Position 10298764 seen before 10298764.
What would be the best way to deal with this without altering fgwas behavior ?

Many thanks
Audrey

'std::bad_alloc'

Hello Joseph!

Thank you so much for your great program! The original publication of fGWAS is a very nice read too.

I am having problems running fGWAS though. I get the following error message:

terminate called after throwing an instance of 'std::bad_alloc'
what(): std::bad_alloc

Do you know what might be causing it? I must warn you that the input file contains ~30M SNPs. When I load a single chromosome only, fGWAS seems to run. I tried increasing -k but didn't seem to solve the problem.

Any help/insight would be much appreciate it!

Best,
Abdou

multithreading

Should be easy to do multithreading of the llk() command using openmp.