multithreadACME: Fast Parallelized Estimation of ACME Model for All Local...

In ACMEeqtl: Estimation of Interpretable eQTL Effect Sizes Using a Log of Linear Model

Description

This function estimates the ACME model (see the vignette for model details) for all gene-SNP pairs within pre-defined distance (cisdist). The input data must be stored in filematrices (see filematrix package) and the results are also saved in a filematrix. This allows the function to perform estimation using multiple CPU cores in parallel without having to duplicate the data across all jobs.

Usage

multithreadACME(
    genefm = "gene",
    snpsfm = "snps",
    glocfm = "gene_loc",
    slocfm = "snps_loc",
    cvrtfm = "cvrt",
    acmefm = "ACME",
    cisdist = 1e+06, 
    threads = -1,
    workdir = ".",
    verbose = TRUE)

Arguments

genefm	Name of the filematrix with gene expression data. One column per gene and one row per sample.
snpsfm	Name of the filematrix with SNP data. One column per SNP and one row per sample.
glocfm	Name of the filematix with gene location information. Must contain two columns, first with gene start location and second with the gene end. The locations must be stored as numbers, the locations for different chromosomes must differ greatly. We suggest encoding (location = 1e9 * chromosome + position_on_chromosome). The rows must match the columns of the genefm filematrix.
slocfm	Name of the filematrix with SNP locations. Must have one column and rows matching columns of snpsfm filematrix. See the instructions for glocfm above.
cvrtfm	Name of the filematirx with covariates. Must not include constant (it is added automatically). One column per covariate and one row per sample.
acmefm	Name of the filematrix to store the estimates. The filemarix will be created. If the filematrix exists, it will be overwritten.
cisdist	The maximum allowed distance between genes and SNPs. Gene-SNP pairs further than cisdist apart will not be tested.
threads	The number of local jobs (CPU cores) used for calculation. If negative, threads is set to the number of cores of the host machine.
workdir	Directory where the input filematrices are located.
verbose	Set to TRUE to indicate progress.

Value

The function creates a filematrix named acmefm with 10 rows and a column for each tested gene-SNP pair. The rows contain gene-SNP ids and the estimates by effectSizeEstimationC:

geneid	The gene id - the column number for the gene in the genefm filematrix.
snp_id	The SNP id - the column number for the SNP in the snpsfm filematrix.
beta0	The constant parameter in the non-linear model.
beta1	The effect size parameter in the non-linear model.
nits	Number of iterations till convergence of the estimation algorithm.
SSE	Sum of squared residuals of the fitted model.
SST	Sum of squared residuals of the model with zero effect.
F	The F test for the significance of the genotype effect.
eta	The effect size parameter for simplified model (beta1/beta0).
SE_eta	Standard error of the eta estimate.

Note

The rows of genefm, snpsfm, and cvrtfm filematrices must match. The SNPs must have increasing locations.

Author(s)

Andrey A Shabalin [e-mail], John Palowitch

References

The manuscript is available at: http://onlinelibrary.wiley.com/doi/10.1111/biom.12810/full

See Also

For package overview and code examples see the package vignette via:
browseVignettes("ACMEeqtl")
or
RShowDoc("doc/ACMEeqtl.html", "html", "ACMEeqtl")

Examples

# First we generate a eQTL dataset in filematrix format 
tempdirectory = tempdir()
z = create_artificial_data(
    nsample = 50,
    ngene = 11,
    nsnp = 51,
    ncvrt = 1,
    minMAF = 0.2,
    saveDir = tempdirectory,
    returnData = FALSE,
    savefmat = TRUE,
    savetxt = FALSE,
    verbose = FALSE)

# In this example, we use 2 CPU cores (threads) 
# for testing of all gene-SNP pairs within 100,000 bp.
multithreadACME(
    genefm = "gene",
    snpsfm = "snps",
    glocfm = "gene_loc",
    slocfm = "snps_loc",
    cvrtfm = "cvrt",
    acmefm = "ACME",
    cisdist = 10e+06,
    threads = 1, # Use more for faster run
    workdir = file.path(tempdirectory, "filematrices"),
    verbose = FALSE)

# Now the filematrix `ACME` holds estimations for all local gene-SNP pairs.

fm = fm.open(file.path(tempdirectory, "filematrices", "ACME"))
TenResults = fm[,1:10]
rownames(TenResults) = rownames(fm)
close(fm)

show(t(TenResults))

ACMEeqtl documentation built on May 2, 2019, 4:03 p.m.