tests/testthat/test_kcone.R

In cdiener/dycone: dycone - analyze the dynamic landscape of metabolome data

# test_kcone.R Copyright 2015 Christian Diener <ch.diener@gmail.com> MIT license.
# See LICENSE for more information.

context("kcone analysis")

test_that("cones can be calculated and analyzed", {
    S <- t(c(1, -1))
    rownames(S) <- "A"
    V <- polytope_basis(S)
    expect_equal(as.vector(V), c(1, 1))
    expect_equal(as.vector(kcone_null(S, c(1, 1))), c(sqrt(2), sqrt(2))/2)
    mats <- runif(2, 0, 10)
    K <- kcone(V, mats)
    expect_equal(dim(K), dim(V))
    expect_equal(K[2]/K[1], mats[1]/mats[2])
    K <- kcone(V, mats, normalize = T)
    expect_equal(K[2]/K[1], mats[1]/mats[2])
    expect_equal(sum(K * K), 1)
    expect_true(inside(rev(mats * 0.42), S, mats))
})

test_that("stability analysis works", {
    S <- matrix(c(1, 0, 0, 1, -1, 0, 0, -1), nrow = 2)
    rownames(S) <- c("A", "B")
    V <- polytope_basis(S)
    stab <- stability_analysis(V, S, c(A = 1, B = 1))
    expect_true(all(stab$what == "stable"))
    expect_equal(stab$ev1, c(0, 0))
    expect_equal(stab$ev2, -0.5 * rep(sqrt(2), 2))
    stab_one <- stability_analysis(V[, 1], S, c(A = 1, B = 1))
    expect_equal(stab_one$ev1, 0)
    expect_equal(stab_one$ev2, -0.5 * sqrt(2))
})

test_that("helper functions work", {
    x <- runif(10)
    x <- x / enorm(x)
    expect_equal(sum(x * x), 1)
    B <- matrix(c(0, 1, 1, 1), ncol = 2)
    class(B) <- "basis"
    expect_equal(occupation(B), c(0.5, 1))
    x <- 1:0
    y <- c(0, 2)
    expect_equal(angle(x, y), pi / 2)
    expect_equal(scaling(y, 0:1), 2)
})

test_that("eigendynamics work", {
    B <- matrix(runif(10000), nrow = 10)
    class(B) <- append(class(B), "basis")
    expect_true(all(eigendynamics(B) >= 0))
    expect_equal(dim(eigendynamics(B, 3)), c(nrow(B), 3))
})

test_that("hypothesis generation works", {
    data(eryth)
    n_r <- length(make_irreversible(eryth))
    mats <- matrix(runif(6 * n_r), ncol = 6)
    sa <- rep(c("normal", "disease"), each = 3)
    h <- hyp(eryth, sa, mats, type = "bias")
    expect_equal(nrow(h), n_r)
    expect_true(all(h$pval <= 1))
    expect_true(all(h$k_lfc == -h$v_lfc))
    o <- c(pyr = -1, prpp = -1, "3pg" = -1, nadph = -1)
    h_opt <- hyp(eryth, sa, mats, type = "fva",
        v_min = 1e-16, obj = o, full = TRUE)
    expect_equal(length(h_opt), 5)
    expect_equal(dim(h_opt$fva), c(n_r, 2))
    expect_equal(length(h_opt$lfc_va), n_r)
    expect_true("fva_log_fold" %in% names(h_opt$hyp))
    fl <- matrix(runif(6 * n_r), ncol = 6)
    h_ex <- hyp(eryth, sa, mats, fl, type = "exact")
    expect_equal(nrow(h_ex), n_r)
    expect_true(all(h_ex$pval <= 1))
    expect_false(all(h_ex$k_lfc == -h_ex$v_lfc))
})

test_that("basis plotting works", {
    B <- matrix(runif(10000), nrow = 10)
    class(B) <- append(class(B), "basis")
    f <- tempfile("plot", fileext = ".png")
    png(f)
    out <- capture.output(plot_basis(B, n_cl = 100))
    dev.off()
    expect_true(file.exists(f))
    expect_match(out, "in-cluster distance", all = F)
    f <- tempfile("plot", fileext = ".png")
    png(f)
    out <- capture.output(plot_red(list(B), n_cl = 100,
        r_names = as.character(1:10)))
    dev.off()
    expect_true(file.exists(f))
    expect_match(out, "Information", all = F)
    expect_match(out, "in-cluster", all = F)
})