Package 'wISAM'

Title: Weighted Inbred Strain Association Mapping
Description: In the course of a genome-wide association study, the situation often arises that some phenotypes are known with greater precision than others. It could be that some individuals are known to harbor more micro-environmental variance than others. In the case of inbred strains of model organisms, it could be the case that more organisms were observed from some strains than others, so the strains with more organisms have better-estimated means. Package 'wISAM' handles this situation by allowing for weighting of each observation according to residual variance. Specifically, the 'weight' parameter to the function conduct_scan() takes the precision of each observation (one over the variance).
Authors: Robert W. Cort <[email protected]>
Maintainer: Robert W. Cort <[email protected]>
License: GPL-3
Version: 0.2.8
Built: 2024-11-21 03:34:10 UTC
Source: https://github.com/rcorty/wisam

Help Index

check_eigen_decomposition

Description

Grabbed from MASS. Useful to sparsify matrices when some eigenvalues are essentially zero.

Usage

check_eigen_decomposition(e, tol = 1e-06)

Arguments

e

the eigen decomposition to check
tol

the threshold below which a number is said to be effectively zero, defaults to 1e-6

Value

The eigen values with any values with absolute value less than tol zeroed.

Example covariate matrix

Description

Example covariate matrix

Usage

covariate_mat

Format

A matrix with 200 rows and 4 variables

calc_multiplier_eigen

Description

Compute a multiplier (aka rotation) matrix. Details in in h2lmm_math_RWC.Rmd.

Value

an object of class GenomeScan

conduct_scan

Description

Conducts the GenomeScan.

Value

an object of class GenomeScan

Note

TODO: allow user to specify subset of chromosomes or loci

Examples

library(wISAM)

wgs <- GenomeScan$new(y = phenotype,
                      X = covariate_mat,
                      G = locus_list,
                      K = kinship_mat,
                      w = 1/se_mean_per_strain)

result <- wgs$prep_scan()$conduct_scan()

fit_locus

Description

Fit one locus of a GenomeScan. Should not typically be called by a user.

Value

an object of class GenomeScan

fit_locus_given_h2

Description

Fit one locus at a specified value of h2. Should not typically be called by a user.

Value

an object of class GenomeScan

initialize

Description

Initialize a GenomeScan

Arguments

y

vector of length n - the phenotype of each of n genomes (individuals or strains)
X

matrix of dimension n-by-c - the covariate value of each individual for c covariates
G

a list where each element is of length n - the genotype of each individual at p loci
K

matrix of dimension n-by-n - the covariance of the phenotype
w

matrix of dimension n-by-n - the inverse variance of the phenotype

Value

an object of class GenomeScan

Examples

library(wISAM)

wgs <- GenomeScan$new(y = phenotype,
                      X = covariate_mat,
                      G = locus_list,
                      K = kinship_mat,
                      w = 1/se_mean_per_strain)

prep_scan

Description

Prepare a GenomeScan for running. Does all the computations that need to be done exactly once per genome scan.

Value

an object of class GenomeScan

Examples

library(wISAM)

wgs <- GenomeScan$new(y = phenotype,
                      X = covariate_mat,
                      G = locus_list,
                      K = kinship_mat,
                      w = 1/se_mean_per_strain)

result <- wgs$prep_scan()

GenomeScan

Description

A Reference Class implementing a Genome Scan

Fields

Options

these are options

Data

Things the user inputs. They have interpretable meaning and define the GenomeScan. Currently: y, X, G, K, weights (inverse variances), and variances.

Intermediates_per_scan

Things the GenomeScan will compute once per scan. They are mathetmatical tools that can't really be interpreted. Currently: L, eigen_L, and LL_null.

Intermediates_per_locus

Things the GenomeScan will compute once per locus. They are mathematical tools that can't really be interpreted. Currently: XG

Intermediates_per_fit

Things the GenomeScan will compute many times per locus (once per trial fit on that locus). These are interpretable but rapidly changing and not guaranteed to be finalized or optimal. Currently: M, LDV, and h2

Results

The results of the GenomeScan. Currently: The h2 that maximizes the LL at each locus and the LR as compared with the no-locus (null) model.

Example kinship matrix

Description

Example kinship matrix

Usage

kinship_mat

Format

A matrix with 200 rows and 200 columns

Example locus list

Description

Example locus list

Usage

locus_list

Format

A list of matrices, where each matrix has 200 rows and 1 column

Example phenotype

Description

Example phenotype

Usage

phenotype

Format

A vector of length 200

Example standard error of the mean per strain

Description

Example standard error of the mean per strain

Usage

se_mean_per_strain

Format

A vector of length 200