Package 'SEAGLE'

Function for applying t(A) on the left for REML EM

Description

Function for applying t(A) on the left for REML EM

Usage

applyAt(qrXtilde, RHS)

Arguments

qrXtilde	Object from QR decomposition of Xtilde
RHS	Object on right hand side of null of Xtilde^T

Value

Matrix or vector resulting from left multiplication of At with matrix or vector input RHS

---

Synthetic haplotype data generated from COSI software

Description

A dataset containing 10,000 haplotypes of SNP sequences mimicking the European population generated from the COSI software

Usage

data(cosihap)

Format

An object of class dgCMatrix with 10000 rows and 604 columns.

Source

https://genome.cshlp.org/content/15/11/1576.abstract

---

REML EM Algorithm

Description

REML EM algorithm for estimating variance components

Usage

estimate.vc(
  y,
  Xtilde,
  qrXtilde,
  beta,
  G,
  init.sigma = 0.5,
  init.tau = 0.5,
  tol = 0.001,
  maxiters = 1000
)

Arguments

y	Vector of observed phenotypes
Xtilde	Matrix of covariates (first column contains the intercept, last column contains the E factor for studying the GxE effect)
qrXtilde	Object containing QR decomposition of Xtilde
beta	Coefficient vector for covariate matrix Xtilde
G	Matrix of genotype markers
init.sigma	Initial sigma input (Default is 0.5)
init.tau	Initial tau input (Default is 0.5)
tol	Tolerance for convergence (Default is 1e-3)
maxiters	Maximum number of iterations (Default is 1000)

Value

Estimates for tau and sigma

---

Generate synthetic data according to a fixed effects model

Description

This function generates synthetic from the fixed effects model described in the experimental studies portion of the paper.

Usage

makeSimData(
  H,
  n,
  L = 100,
  maf = 0.01,
  gamma0 = 1,
  gammaX = 1,
  gammaE = 1,
  gammaG,
  gammaGE,
  causal = 40,
  seed = 12345
)

Arguments

H	Matrix of haplotype data (e.g. cosihap)
n	Number of individuals
L	Number of SNPs in the G matrix (Default is 100), should be a value between 1 and 604
maf	Minor allele frequency (Default is 0.01)
gamma0	gamma0 Fixed effect coefficient for intercept (Default is 1)
gammaX	gammaX Fixed effect coefficient for confounding covariates (Default is 1)
gammaE	gammaE Fixed effect coefficient for E effect (Default is 1)
gammaG	gammaG Fixed effect coefficient for G main effect
gammaGE	gammaGE Fixed effect coefficient for GxE interaction effect
causal	Number of causal SNPs (default is 40)
seed	Seed (Default is 12345)

Value

Synthetic dataset containing y, X, E, G, epsilon, and number of causal SNPs

Examples

dat <- makeSimData(H=cosihap, n=500, L=10, gammaG=1, gammaGE=0, causal=4, seed=1)

---

Prepare data for input into SEAGLE function

Description

This function checks and formats data for input into SEAGLE function

Usage

prep.SEAGLE(y, X, intercept, E, G)

Arguments

y	Vector of observed phenotypes
X	Matrix of covariates without genetic marker interactions
intercept	1 if the first column of X is the all ones vector, 0 otherwise
E	E Vector of environmental covariates
G	G Matrix of genotype data

Value

List object containing prepared data for input into SEAGLE function

Examples

dat <- makeSimData(H=cosihap, n=500, L=10, gammaG=1, gammaGE=0, causal=4, seed=1)
objSEAGLE <- prep.SEAGLE(y=dat$y, X=dat$X, intercept=1, E=dat$E, G=dat$G)

---

Function for applying R inverse to AtG in REML EM algorithm

Description

Function for applying R inverse to AtG in REML EM algorithm

Usage

Rinv.AtG(G, AtG, GtAAtG, tau, sigma)

Arguments

G	Matrix of genotype markers (size n x L)
AtG	AtG from pre-computation
GtAAtG	GtAAtG from pre-computation
tau	Variance component from G main effect
sigma	Variance component from model noise epsilon

Value

Matrix resulting from left multiplication of Rinv with input matrix AtG

---

Function for applying R inverse to u in REML EM algorithm

Description

Function for applying R inverse to u in REML EM algorithm

Usage

Rinv.u(G, AtG, GtAAtG, GtAu, u, tau, sigma)

Arguments

G	Matrix of genotype markers (size n x L)
AtG	AtG from precomputation
GtAAtG	GtAAtG from precomputation
GtAu	GtAu from precomputation
u	u=Aty from REML EM
tau	Variance component from G main effect
sigma	Variance component from model noise epsilon

Value

Vector resulting from left multiplication of Rinv with input vector u

---

Compute score-like test statistic and p-value for GxE test with SEAGLE algorithm

Description

This function computes the score test statistic and corresponding p-value for the GxE test with the SEAGLE algorithm with input data that have been prepared with the prep.SEAGLE function

Usage

SEAGLE(obj.SEAGLE, init.tau = 0.5, init.sigma = 0.5, pv = "liu")

Arguments

obj.SEAGLE	Input data prepared with prep.SEAGLE function
init.tau	Initial estimate for tau (Default is 0.5)
init.sigma	Initial estimate for sigma (Default is 0.5)
pv	Method of obtaining p-value (Either "liu" or "davies", Default is liu)

Value

Score-like test statistic T for the GxE effect and corresponding p-value

Examples

dat <- makeSimData(H=cosihap, n=500, L=10, gammaG=1, gammaGE=0, causal=4, seed=1)
objSEAGLE <- prep.SEAGLE(y=dat$y, X=dat$X, intercept=1, E=dat$E, G=dat$G)
res <- SEAGLE(objSEAGLE, init.tau=0.5, init.sigma=0.5)

---

Function for applying V inverse in Algorithm 1

Description

This function applies V inverse via the Woodbury matrix identity

Usage

Vinv(G, qrM, tau_over_sigma, sigma, RHS)

Arguments

G	Matrix of genotype markers (size n x L)
qrM	Pre-computation for LxL linear system solve
tau_over_sigma	Tau over sigma from precomputation
sigma	Variance component from model noise epsilon
RHS	Matrix or vector on right-hand side of V inverse

Value

Matrix or vector resulting from left multiplication of Vinv with input RHS

---