binQTLScan.R

binQTLScan <- function(phenotype="", genotype="", population="RIL") {
  if (population == "MAGIC") {
    result <- magicbinQTL(phenotype=phenotype, genotype=genotype)
    
    return(result)
  } else {
    Z <- as.matrix(genotype[, -c(1:4)])
    Y <- as.matrix(phenotype[, 2]) 
    genotype.n <- genoConv(genotype)
    kk <- kinship(genotype.n$genotype.a)
    m <- nrow(Z)
    n <- ncol(Z)
    X <- as.matrix(rep(1, n))
    
    qq <- eigen(kk, symmetric = T)
    uu <- qq$vectors
    delta <- qq$values
    
    r <- ncol(X)
    x <- t(uu) %*% X
    y <- t(uu) %*% Y
    if (population=="RIL") {
      L <- matrix(c(-0.5, 0.5), 1, 2)
    } else {
      L <- rbind(c(-0.5, 0, 0.5), c(-0.5, 1, -0.5))
    }
    invLL <- solve(L %*% t(L))
    rnk <- nrow(L)
    
    triplet <- TRUE
    parr <- numeric()
    for (k in 1:m) {
      if ((triplet == FALSE) | (k == 1) | (k == m)) {
        U <- as.factor(Z[k, ])
        Zk <- model.matrix( ~ U - 1)
        zk <- t(uu) %*% as.matrix(Zk)
        fit <- random( x = x, z = zk, y = y, m = m, kk = qq, cov = "qq", delta = delta, n = n)
      } else{
        U1 <- as.factor(Z[k - 1, ])
        U2 <- as.factor(Z[k, ])
        U3 <- as.factor(Z[k + 1, ])
        Z.1 <- model.matrix( ~ U1 - 1)
        Z.2 <- model.matrix( ~ U2 - 1)
        Z.3 <- model.matrix( ~ U3 - 1)
        Zk <- cbind(Z.1, Z.2, Z.3)
        zk <- t(uu) %*% as.matrix(Zk)
        fit <- random1( x = x, z = zk, y = y, m = m, kk = qq, cov = "qq", delta = delta, n = n )
      }
      covparm <- fit[[1]]
      vg <- covparm[1]
      vk <- covparm[2]
      ve <- covparm[3]
      lambda <- covparm[4]
      lambda.k <- covparm[5]
      beta <- fit[[2]]
      gamma <- fit[[3]]
      stderr <- fit[[4]]
      cov <- fit[[5]]
      lrt <- fit[[6]]
      effect <- L %*% gamma
      vv <- L %*% cov %*% t(L)
      wald <- t(effect) %*% solve(vv) %*% effect
      dk <- rnk - sum(diag(invLL %*% vv)) / vk
      if (dk < 1e-5) { pwald <- 1 } else { pwald <- 1 - pchisq(wald, dk) }
      if (lrt < 1e-8) { plrt <- 1 } else { plrt <- 0.5 * (1 - pchisq(lrt, 1)) }
      par <- c( k, vg, vk, ve, lambda, lambda.k, drop(beta), drop(gamma), drop(stderr),
                drop(effect), wald, dk, pwald, -log10(pwald), lrt, plrt, -log10(plrt)  )
      parr <- rbind(parr, par)
    }
    
    if (population=="RIL") {
      varnames <- c("k", "vg", "vk", "ve", "lambda", "lambdak", "beta", "g11",
                    "g22", "sd11", "sd22")
      varnames <- c(varnames, "additive", "wald", "dk", "pwald", "logpwald",
                    "lrt", "plrt", "logplrt")
    } else if (population=="F2") {
      varnames <- c("k", "vg", "vk", "ve", "lambda", "lambdak", "beta", "g11",
                    "g12", "g22", "sd11", "sd12", "sd22")
      varnames <- c(varnames, "additive", "dominance", "wald", "dk", "pwald",
                    "logpwald", "lrt", "plrt", "logplrt")
    }
    
    out <- as.matrix(parr)
    colnames(out) <- varnames
    rownames(out) <- 1:nrow(out)
    result <- data.frame(genotype[, 1:4], out, stringsAsFactors=FALSE)
    result <- result[, c("Bin", "Chr", "Start", "Stop", "plrt", "logplrt")]
    names(result)[5:6] <- c("p", "-logp")
    
    return(result)
  }
}