soedinglab/b-lore

Name: b-lore

Owner: Söding Lab

Description: Bayesian multiple logistic regression for GWAS meta-analysis

Created: 2017-07-06 14:16:22.0

Updated: 2017-11-01 09:00:52.0

Pushed: 2017-10-17 12:59:32.0

Homepage: null

Size: 21519

Language: Python

GitHub Committers

User	Most Recent Commit	# Commits

Other Committers

User	Email	Most Recent Commit	# Commits

README

B-LORE

Bayesian LOgistic REgression

A tool for meta-analysis in GWAS using Bayesian multiple logistic regression

Description

B-LORE is a command line tool that creates summary statistics from multiple logistic regression on GWAS data, and combines the summary statistics from multiple studies in a meta-analysis. It can also incorporate functional information about the SNPs from other external sources. Several genetic regions, or loci are preselected for analysis with B-LORE.

Key features

1. Association probability: B-LORE outputs probabilities of the input genetic loci being statistically associated with the phenotype.
2. Finemapping: B-LORE also outputs the probability of each SNP being statistically associated with the phenotype.
3. Leverage functional genomic data as a prior probability to improve prioritization.
4. Models data with logistic regression, and is suited for case/control studies.
5. Combines information over all SNPs in a locus with multiple regression.

Installation

B-LORE is written in python and C++. To run B-LORE, you will need

• python version 3.4 or higher,
• the Python packages for scientific computing NumPy and SciPy.
• C++ compiler

To use B-LORE, you have to download the repository and compile the C++ shared libraries:

clone https://github.com/soedinglab/b-lore.git
-lore

The Makefile uses g++ by default, which you can change depending on the compiler available on your system.

Input files

For calculating summary statistics, it uses the following file formats as input:

1. Genotype files in Oxford format, for all loci of interest.
2. Sample file in Oxford format

For meta-analysis, it uses the following input:

1. Output from B-LORE summary statistics. Note that it cannot use standard SNPTEST summary statistics.
2. (Optional) Functional genomics data, separately for each locus. Each feature file contains 2 parts: (a) a header line detailing the names of the columns in the file, and (b) a line for each SNP detailing the information for that SNP. The columns are tab-separated. The annotation tracks are present from column 4 onwards. The first 3 columns are:
   • RSID: must have the same SNP identifier as in the genotype files
   • CHR: chromosome number
   • POS: base-pair position of the SNP.

Usage

Quick start

• Clone the repository
• cd example
• tar -zxvf input.tar.gz This will create an example input folder, with genotypes at 20 loci for 3 populations, a sample file for each population and ENCODE data for the 20 loci.
• ./commands.sh to run B-LORE on the 3 populations to generate summary statistics, followed by a meta-analysis.

Command line arguments

An executable file to run B-LORE is provided as bin/blore. This can used as follows:

e [--help] [COMMAND] [OPTIONS]

There are 2 commands for B-LORE:

• --summary : for creating summary statistics of individual studies.
• --meta : for meta-analysis from summary statistics of multiple studies.

Each of these 2 commands takes different options, as described below.

blore –summary [OPTIONS]

Create summary statistics of individual studies. Valid options are:

Option | Description | Priority | Default value :— | :— |:— | :– ‑‑gen filename(s) | Input genotype file(s), all loci should have separate genotype files and specified here (wildcards allowed) | Required | – ‑‑sample filename | Input sample file | Required | – ‑‑pheno string | Name of the phenotype as it appears in the header of the sample file| Optional | pheno ‑‑regoptiom | If specified, the variance of the regularizer will be optimized, otherwise it will be N(0, ?²) where ? is specified by --reg | Optional | – ‑‑reg float | Value of the standard deviation (?) of the regularizer | Optional | 0.01 ‑‑pca int | Number of principal components of the genotype to be included as covariates | Optional | 0 ‑‑cov string(s) | Name of covariate(s) as they appears in the header of the sample file, multiple covariates can be specified as space-separated strings | Optional | None ‑‑out directory | Name of the output directory where summary statistics will be created | Optional | directory of the genotype files ‑‑prefix string | Prefix for the summary statistics files | Optional | _summary

blore –meta [OPTIONS]

Perform meta-analysis from summary statistics of multiple studies. Valid options are:

Option | Description | Priority | Default value :— | :— |:— | :– ‑‑statinfo filename(s) | Input file prefix(es) of summary statistics, full path is required | Required | – ‑‑feature filename(s) | Input file(s) for genomic feature tracks | Optional | – ‑‑params floats | Initial values of the hyperparameters, requires 4 space-separated floats corresponding to ?_? ? ? ?_bg| Optional | 0.01 0.0 0.01 0.01 ‑‑muvar | If specified, ? will be optimized, otherwise it will be fixed to the initial value | Optional | – ‑‑zmax int | Maximum number of causal SNPs allowed | Optional | 2 ‑‑out directory | Name of the output directory where result files will be created | Optional | current directory ‑‑prefix string | Prefix for the meta-analysis output files | Optional | _meta

Example

• Clone the repository
• cd example
• tar -zxvf input.tar.gz This will create an example input folder, with genotypes at 20 loci for 3 populations, a sample file for each population and ENCODE data for the 20 loci.

View commands.sh in your favorite editor to see the commands, and execute ./commands.sh to run B-LORE on the 3 populations to generate summary statistics, followed by a meta-analysis.

Citation

• Saikat Banerjee, Lingyao Zeng, Heribert Schunkert and Johannes Soeding (2017). Bayesian multiple logistic regression for meta-analyses of GWAS. bioRxiv.

License

B-LORE is released under the GNU General Public License version 3. See LICENSE for more details. Copyright Johannes Soeding and Saikat Banerjee.

Contact

Saikat Banerjee