HASE

Framework for efficient high-dimensional association analyses.

Speed test

run_ExampleStudy.sh script runs example of association study with 20.000 SNPs, 1000 phenotypes and 1000 subjects. It runs analysis by chunk of 5000 SNPs (which you can define in config.py file). Standard output looks like this:

START regression mode...
reading file example_study.csv
There are 1000 ids and 1000 columns 
reading file example_study.csv
There are 1000 ids and 3 columns 
There are 1000 ids
There are 1000 common ids
...
...
...
time to compute GWAS for 1000 phenotypes and 5000 SNPs .... 0.681949138641 sec
Read 15000, processed 15000, total 20000
...
time to compute GWAS for 1000 phenotypes and 5000 SNPs .... 0.565479040146 sec
Read 20000, processed 20000, total 20000
...
experiment finished in 10.0326929092 s

Installation HASE

Navigate to directory where you want to install HASE and clone this repository: git clone https://github.com/roshchupkin/hase.git

Update HASE

You can update HASE to newest version using git. Navigate to your HASE folder (where you cloned git repository):
git pull

Installation requirements

Your system might already satisfied requirements, we suggest first try to run test example from Testing header below.

1. HDF5 software (python packages tables and h5py require this installation). If it is not installed on you system, you can download to your home directory the latest source code hdf5.

   tar -xf ~/hdf5-1.8.16.tar.gz
   cd ~/hdf5-1.8.16/
   ./configure 
   make 
   make install
   

   Then you need to add one line to your .bachrc or .bash_profile file in your home directory.

   export HDF5_DIR=~/hdf5-1.8.16/hdf5/
   

2. BLAS and LAPACK linear algebra libraries for scipy and numpy.

   sudo apt-get install gfortran libopenblas-dev liblapack-dev
   

   If this does not work or raise errors, then you might need to follow instruction from scipy website.

3. You need to install python. You can download python from official website python or install one of the python distribution for scientific research, such as Anaconda, Enthought Canopy or Python(x,y). And then you need to install (or first uninstall) scipy and numpy python libraries.

   pip install scipy 
   pip install numpy
   

   To check linkage in numpy:

   python
   >>> import numpy as np
   >>> np.__config__.show()
   

   And you should see something like this:

   lapack_opt_info:
     libraries = ['openblas', 'openblas']
     library_dirs = ['/cm/shared/apps/openblas/0.2.9-rc2/lib']
     language = f77
   blas_opt_info:
     libraries = ['openblas', 'openblas']
     library_dirs = ['/cm/shared/apps/openblas/0.2.9-rc2/lib']
     language = f77
   openblas_info:
     libraries = ['openblas', 'openblas']
     library_dirs = ['/cm/shared/apps/openblas/0.2.9-rc2/lib']
     language = f77
   blas_mkl_info:
   NOT AVAILABLE
   

4. Install python packages listed in requirements.txt file. (you can use package manager which comes with your python pip or conda to install packages):

   • bitarray
   • argparse
   • cython
   • matplotlib
   • scipy
   • numpy
   • pandas
   • h5py
   • tables

Testing

1. Navigate to HASE directory and type python hase.py -h, you should see help message.
2. Navigate to HASE directory and type sh run_ExampleStudy.sh, it should start running toy example of high-dimensional GWAS.

User Guide

wiki.

Requirements

1. HDF5 software.
2. BLAS and LAPACK linear algebra libraries.
3. Python.
4. Python packages:
   • bitarray
   • argparse
   • cython
   • matplotlib
   • scipy
   • numpy
   • pandas
   • h5py
   • tables
5. Git.

Citation

If you use HASE framework, please cite:

Roshchupkin, G. V. et al. HASE: Framework for efficient high-dimensional association analyses. Sci. Rep. 6, 36076; doi: 10.1038/srep36076 (2016)

Licence

This project is licensed under Apache-2.0 License.

Authors

Gennady V. Roshchupkin (Department of Epidemiology, Radiology and Medical Informatics, Erasmus MC, Rotterdam, Netherlands)

Hieab H. Adams (Department of Epidemiology, Erasmus MC, Rotterdam, Netherlands)

Contacts

If you have any questions/suggestions/comments or problems do not hesitate to contact us!

• [email protected]
• [email protected]