tests/testthat/test-reduceFeatures.R

In HimesGroup/qmtools: Quantitative Metabolomics Data Processing Tools

test_that("reduceFeatures works.", {
    ## PCA with missing
    m <- assay(faahko_sub, i = "raw")
    out1 <- reduceFeatures(faahko_sub, i = "raw", name = "pca", method = "pca")
    expect_s3_class(out1, "reduced.pca")
    out2 <- reduceFeatures(m, method = "pca")
    expect_identical(out1, out2)

    ## PCA
    m <- assay(faahko_sub, i = "knn_vsn")
    out1 <- reduceFeatures(faahko_sub, i = "knn_vsn", name = "pca",
                           method = "pca", ncomp = 3)
    expect_s3_class(out1, "reduced.pca")
    out2 <- reduceFeatures(faahko_sub, i = "knn_vsn", name = "pca",
                           method = "pca", ncomp = 2)
    out3 <- reduceFeatures(m, method = "pca", ncomp = 2)
    expect_false(identical(dim(out1), dim(out3)))
    expect_identical(out2, out3)

    ## t-SNE
    m <- assay(faahko_sub, i = "knn_vsn")
    set.seed(1e7)
    out1 <- reduceFeatures(faahko_sub, i = "knn_vsn", name = "tsne",
                           method = "tsne", perplexity = 3)
    expect_s3_class(out1, "reduced.tsne")
    set.seed(1e7)
    out2 <- reduceFeatures(m, method = "tsne", perplexity = 3)
    expect_identical(out1, out2)
    expect_error(
        reduceFeatures(faahko_sub, i = "raw", method = "tsne", perplexity = 3)
    )

    ## PLS-DA
    y <- factor(colData(faahko_sub)$sample_group)
    m <- assay(faahko_sub, i = "knn_vsn")
    out1 <- reduceFeatures(faahko_sub, i = "knn_vsn", name = "tsne",
                           method = "plsda", y = y)
    expect_s3_class(out1, "reduced.plsda")
    out2 <- reduceFeatures(m, method = "plsda", y = y)
    expect_identical(out1, out2)
    expect_error(
        reduceFeatures(faahko_sub, i = "raw", method = "plsda", y = y) # missing
    )
    expect_error(
        reduceFeatures(faahko_sub, i = "raw", method = "plsda",
                       y = colData(faahko_sub)$sample_group) # not factor
    )
})