Add tests for summariseByExon

Covers end-to-end execution on the bundled fixture SE and verifies the
exon-aggregation identity against synthetic Poisson counts, including a
per-exon hand-check on the most-shared exon.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: cying111

tests/testthat/test_summariseByExon.R

@@ -0,0 +1,57 @@
+context("Exon quantification")
+
+test_that("summariseByExon runs on bundled SE and returns a well-formed SE", {
+    se <- readRDS(system.file("extdata",
+        "seOutput_SGNex_A549_directRNA_replicate5_run1_chr9_1_1000000.rds",
+        package = "bambu"))
+
+    seExon <- summariseByExon(se)
+
+    expect_s4_class(seExon, "RangedSummarizedExperiment")
+    expect_true("counts" %in% assayNames(seExon))
+    expect_equal(ncol(seExon), ncol(se))
+    expect_equal(colnames(seExon), colnames(se))
+    expect_true("GENEID" %in% colnames(mcols(rowRanges(seExon))))
+    # rownames are unique chr:start:end:strand keys
+    expect_equal(length(unique(rownames(seExon))), nrow(seExon))
+})
+
+test_that("summariseByExon aggregation identity holds with synthetic counts", {
+    se <- readRDS(system.file("extdata",
+        "seOutput_SGNex_A549_directRNA_replicate5_run1_chr9_1_1000000.rds",
+        package = "bambu"))
+
+    # Rebuild SE with two samples of Poisson counts (bundled fixture is all zero)
+    set.seed(1)
+    n <- nrow(se)
+    cnt <- Matrix::Matrix(cbind(
+        sampleA = as.numeric(rpois(n, 5)),
+        sampleB = as.numeric(rpois(n, 10))
+    ), sparse = TRUE)
+    rownames(cnt) <- rownames(se)
+    rr <- rowRanges(se)
+    mcols(rr) <- rowData(se)
+    se2 <- SummarizedExperiment(
+        assays    = S4Vectors::SimpleList(counts = cnt),
+        rowRanges = rr,
+        colData   = DataFrame(id = c("sampleA", "sampleB"),
+                              row.names = c("sampleA", "sampleB")))
+
+    seExon <- summariseByExon(se2)
+
+    # Per-sample identity: colSums(exon) == sum_tx #exons(tx) * count(tx)
+    exonPerTx <- lengths(rowRanges(se2))
+    expected <- colSums(as.matrix(cnt) * exonPerTx)
+    observed <- colSums(as.matrix(assays(seExon)$counts))
+    expect_equal(unname(observed), unname(expected))
+
+    # Spot-check the most-shared exon against a manual tx-set sum
+    ex  <- unlist(rowRanges(se2), use.names = TRUE)
+    key <- paste(as.character(seqnames(ex)), start(ex), end(ex),
+                 as.character(strand(ex)), sep = ":")
+    top <- names(sort(table(key), decreasing = TRUE))[1]
+    txs <- unique(names(ex)[key == top])
+    manual <- colSums(as.matrix(cnt[txs, , drop = FALSE]))
+    expect_equal(unname(as.numeric(assays(seExon)$counts[top, ])),
+                 unname(manual))
+})