RL-SKAT

RL-SKAT is a method for the calculation of exact p-values for the score test in heritability, in the case of a single kernel and a continuous phenotype. RL-SKAT is described in the following paper.

Code written by Regev Schweiger. For questions or comments, mail schweiger@post.tau.ac.il.

Installation

You can simply download the code:

1. Download the ZIP file of the latest release here: https://github.com/cozygene/RL-SKAT/releases.
2. Extract the ZIP file to a folder

RL-SKAT is implemented as a set of Python classes. To use, simply import/execfile (or just %run in IPython) the file, e.g.:

from rl_skat import *

RL-SKAT requires NumPy, SciPy and FaST-LMM (used for the implementation of the Davies method used in the code).

Usage

An object is constructed which does all the required preprocessing, and can then use it to test multiple phenotypes. All matrices should have the same number of rows (n, the sample size).

The file example.py contains a short example (for the heritability use-case), along with example files.

Full kernel

When you have a kernel matrix already computed, use this version:

RL_SKAT_Full_Kernel(kernel_matrix, fixed_covariates=None, add_intercept=True).test(phenotypes, acc=1e-7)

Where:

• kernel_matrix - An n by n kernel matrix.
• fixed_covariates - A set of n by p optional additional covariates with fixed effects.
• add_intercept - Set to True if an intercept should be added as a covariate.
• phenotypes - A matrix of n by N of N phenotypes to test.
• acc - Optional parameter of p-value accuracy. More accurate p-values take longer to calculate.

The method returns N p-values.

Set testing

If you have a set of covariates to be treated as having random effects (that is, the square of the kernel matrix), use this (add covariates as needed):

RL_SKAT_Low_Rank(weighted_Z, fixed_covariates=None, add_intercept=True).test(phenotypes, acc=1e-7)

Where:

• weighted_Z - An n by m matrix of (possibly weighted) covariates corresponding to random effects, e.g. rare variants.
• fixed_covariates - A set of n by p optional additional covariates with fixed effects.
• add_intercept - Set to True if an intercept should be added as a covariate.
• phenotypes - A matrix of n by N of N phenotypes to test.
• acc - Optional parameter of p-value accuracy. More accurate p-values take longer to calculate.

The method returns N p-values.