PolyFun (POLYgenic FUNctionally-informed fine-mapping)
PolyLoc (POLYgenic LOCalization of complex trait heritability)
This page contains the code of the methods PolyFun for functionally-informed fine-mapping and PolyLoc for polygenic localization of complex trait heritability, described in Weissbrod et al. 2019 bioRxiv.
PolyFun estimates prior causal probabilities for SNPs, which can then be used by fine-mapping methods like SuSiE or FINEMAP. Unlike previous methods for functionally-informed fine-mapping, PolyFun can aggregate polygenic data from across the entire genome and hundreds of functional annotations.
PolyLoc generalizes fine-mapping by constructing minimal sets of SNPs that causally explain a given proportion (e.g. 50%) of SNP heritability.
We also provide a script called finemapper that facilitates fine-mapping with methods like SuSiE, saving many of the preprocessing steps often required to perform fine-mapping (e.g. handling allelic flips between the summary statistics and reference genotypes).
The files in the ldstore directory are an adaptation of the ldstore package (written by Christian Benner) to Python 3.
We provide a detailed manual of PolyFun, PolyLoc and finemapper in the Wiki page. If you run into any issues, please check the FAQ first.
We provide several installation options.
The easiest way to install polyfun is by creating a dedicated environment through the Anaconda Python distribution. To do this, please install Anaconda on your machine and then type the following commands:
git clone https://github.com/omerwe/polyfun
cd polyfun
conda env create -f polyfun.yml
conda activate polyfun
This will install all the dependencies except for SuSiE, FINEMAP, and LDstore You can use PolyFun without these packages to compute prior causal probabilities, but you won't be able to apply the actual fine-mapping. Please see installation instructions for these three packages below.
After the installation, you can always invoke the PolyFun environment with the command conda activate polyfun.
We recommend that you frequently make sure you have the latest version of polyfun installed by going to the polyfun directory and typing git pull.
PolyFun and PolyLoc are designed for Python >=3.6 and require the following freely available Python packages:
- numpy and scipy
- scikit-learn
- pandas (version >=0.25.0)
- tqdm
- pyarrow
- bitarray
- networkx (only required for HESS-based estimation of effect size variance)
- pandas-plink
It is recommended (but not required) to also install the following:
- rpy2 (a Python package)
- R version 3.5.1 or higher
- Ckmeans.1d.dp (a package for R, that will be invoked from python via the rpy2 package).
If rpy2 or Ckmeans.1d.dp are not installed, PolyFun and PolyLoc will fallback to suboptimal clustering via scikit-learn.
The finemapper script also requires the following:
- A fine-mapping package you'd like to use. At the moment we support susieR and FINEMAP v1.4. Please see installation instructions for these packages below.
- (optional) The program LDstore 2.0 for computing LD directly from .bgen files (imputed genotypes)
We recommend running PolyFun/PolyLoc via the Anaconda Python distribution. In Anaconda, you can install all the Python packages with the command "conda install <package_name>". Alternatively, the Python packages can be installed with the command "pip install --user <package_name>".
Once all the prerequisite packages are installed, you can install PolyFun/PolyLoc on a git-enabled machine by typing:
git clone https://github.com/omerwe/polyfun
We recommend that you frequently make sure you have the latest version of polyfun installed by going to the polyfun directory and typing git pull.
To install SuSiE, please start an R shell (usually by typing R) and then type:
devtools::install_github("stephenslab/susieR@0.8.0",build_vignettes=FALSE)
If this doesn't work, please refer to the SuSiE website for more information, or contact the SuSiE authors through the SuSiE Github page.
To install FINEMAP v1.4, please type one of the following two commands:
If you use Linux:
wget http://www.christianbenner.com/finemap_v1.4_x86_64.tgz
tar xvf finemap_v1.4_x86_64.tgz
If you use Mac OS X :
wget http://www.christianbenner.com/finemap_v1.4_MacOSX.tgz
tar xvf finemap_v1.4_MacOSX.tgz
To install LDstore, please type one of the following two commands:
If you use Linux:
wget http://www.christianbenner.com/ldstore_v2.0_x86_64.tgz
tar xzvf ldstore_v2.0_x86_64.tgz
If you use Mac OS X :
wget http://www.christianbenner.com/ldstore_v2.0_MacOSX.tgz
tar xzvf ldstore_v2.0_MacOSX.tgz
We recommend testing PolyFun by invoking the script:
python test_polyfun.py --python3 <python3_exe>
where python3_exe (optional) is the command you type to start a python3 session (default is python). If the script completes without an error, everything is fine. If you see any errors, please consult the FAQ.
For questions and comments, please open a Github issue (preferred) or contact Omer Weissbrod at oweissbrod[at]hsph.harvard.edu