fix: resolve transitive BLQ imputation dependencies for half.life chain (#1057)

.gitignore

@@ -15,4 +15,4 @@ inst/shiny/log/*
 .DS_Store
 # {shinytest2}: Ignore new debug snapshots for `$expect_values()`
 *_.new.png
-desktop.ini
+desktop.ini

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: aNCA
 Title: (Pre-)Clinical NCA in a Dynamic Shiny App
-Version: 0.1.0.9175
+Version: 0.1.0.9176
 Authors@R: c(
     person("Ercan", "Suekuer", email = "ercan.suekuer@roche.com", role = "aut",
            comment = c(ORCID = "0009-0001-1626-1526")),

NEWS.md

@@ -18,6 +18,9 @@
 * Partial interval parameters section supports calculations beyond `AUCINT`: `RCAMINT`, `AUCINTD`, `CAVGINT`, and others. Table starts empty by default with a Remove Row button (#524, #1249)
 * "Min. Points for Half-life" setting added (range 2–10, default 3) (#1155)
 * BLQ imputation rules via `NCA Setup > Data Imputation` (#139)
+
+### Bug Fixes
+* Half-life (LAMZHL) and lambda.z (LAMZ) now correctly retain BLQ imputation when half-life-dependent parameters (AUCIFO, AUCIFP) are requested. Previously `rm_impute_obs_params()` removed imputation from `half.life` because its dependency check was limited to one level — missing the transitive chain `half.life -> lambda.z -> aucinf.obs`. The dependency resolution now recursively traverses upstream dependencies to include all parameters in the AUC calculation chain (#1057).
 * General Exclusions section for in-app NCA exclusions, with "Excl. TLG" checkbox per entry (#851, #1018)
 * Parameter Exclusions tab: exclude individual PK parameter rows from descriptive statistics and ADPP export via PPSUMFL/PPSUMRSN flags (#1040)
 * NCA flag rules (NCAwXRS) from ADNCA standards — flagged records are excluded from NCA (#752)

R/intervals.R

@@ -295,6 +295,12 @@ update_main_intervals <- function(
   # and apply it only for non-observational parameters
 
   if (!is.null(blq_imputation_rule)) {
+    # Ensure impute column exists so dplyr mutate below references the
+    # column rather than the function parameter (which could be FALSE
+    # from YAML settings, causing "PKNCA_impute_method_FALSE" error).
+    if (!"impute" %in% names(data$intervals)) {
+      data$intervals$impute <- NA_character_
+    }
     data$intervals <- data$intervals %>%
       mutate(
         impute = ifelse(
@@ -324,16 +330,20 @@ update_main_intervals <- function(
 #' @import dplyr
 #'
 rm_impute_obs_params <- function(data, metadata_nca_parameters = metadata_nca_parameters) {
-  # Don't impute parameters that are not AUC dependent
+  # Don't impute parameters that are not AUC dependent.
+  # Parameters in the AUC calculation chain (half.life feeds lambda.z,
+  # which feeds aucinf.obs) must all use the same BLQ-imputed data (#1057).
   params_auc_dep <- metadata_nca_parameters %>%
     filter(grepl("auc|aumc", PKNCA) | grepl("auc", Depends)) %>%
     pull(PKNCA)
 
+  # Build consumer map ("who consumes X") and walk upstream from AUC params
+  # to find all parameters in their dependency chain (#1057).
+  consumer_map <- .build_consumer_map(metadata_nca_parameters)
+  needs_impute <- .find_upstream_deps(params_auc_dep, consumer_map)
+
   params_not_to_impute <- metadata_nca_parameters %>%
-    filter(
-      !grepl("auc|aumc", PKNCA),
-      !grepl(paste0(params_auc_dep, collapse = "|"), Depends)
-    ) %>%
+    filter(!PKNCA %in% needs_impute) %>%
     pull(PKNCA) %>%
     intersect(names(PKNCA::get.interval.cols()))
 
@@ -367,3 +377,51 @@ rm_impute_obs_params <- function(data, metadata_nca_parameters = metadata_nca_pa
 
   data
 }
+
+#' Build a consumer map from the Depends column.
+#' For each parameter A, returns which parameters consume A (list A as a dependency).
+#' e.g., lambda.z lists half.life in Depends, so consumer_map$half.life = "lambda.z".
+#' @noRd
+.build_consumer_map <- function(metadata) {
+  rev <- list()
+  for (i in seq_len(nrow(metadata))) {
+    pkg_name <- metadata$PKNCA[i]
+    dep_str <- metadata$Depends[i]
+    if (is.na(dep_str) || dep_str == "") next
+    dep_list <- trimws(strsplit(dep_str, ",")[[1]])
+    for (d in dep_list) {
+      rev[[d]] <- unique(c(rev[[d]], pkg_name))
+    }
+  }
+  rev
+}
+
+#' Find all upstream dependencies transitively from `start_set`.
+#' Walks the consumer map to collect params that feed into the current chain,
+#' stopping at purely observational leaf params (cmax, tmax, tlast).
+#'
+#' @param start_set Character vector of starting PKNCA parameter names.
+#' @param consumer_map Named list from `.build_consumer_map()`.
+#' @param obs_params Character vector of observational params to exclude.
+#'   Must be kept in sync with metadata_nca_parameters when new leaf params
+#'   are added. Default: cmax, tmax, tlast.
+#' @param max_iter Maximum iterations to guard against infinite loops from
+#'   circular dependencies. 50 is generous (~40 params in real metadata).
+#' @noRd
+.find_upstream_deps <- function(start_set, consumer_map,
+                                 obs_params = c("cmax", "tmax", "tlast"),
+                                 max_iter = 50L) {
+  needs <- start_set
+  for (iter in seq_len(max_iter)) {
+    newly_found <- character()
+    for (pkg in names(consumer_map)) {
+      if (!pkg %in% needs && !pkg %in% obs_params &&
+          any(consumer_map[[pkg]] %in% needs)) {
+        newly_found <- c(newly_found, pkg)
+      }
+    }
+    if (length(newly_found) == 0) break
+    needs <- c(needs, newly_found)
+  }
+  needs
+}

tests/testthat/test-intervals.R

@@ -449,3 +449,80 @@ describe("rm_impute_obs_params", {
     expect_equal(result$intervals, intervals_before)
   })
 })
+
+describe(".build_consumer_map", {
+  it("builds consumer map from Depends column", {
+    metadata <- data.frame(
+      PKNCA = c("half.life", "lambda.z", "aucinf.obs", "cmax"),
+      Depends = c("tmax, tlast", "half.life", "lambda.z", NA),
+      stringsAsFactors = FALSE
+    )
+    result <- .build_consumer_map(metadata)
+    expect_type(result, "list")
+    expect_true("lambda.z" %in% result[["half.life"]])
+    expect_true("aucinf.obs" %in% result[["lambda.z"]])
+    expect_null(result[["cmax"]])
+  })
+
+  it("handles empty Depends strings", {
+    metadata <- data.frame(
+      PKNCA = c("param1", "param2"),
+      Depends = c("", NA),
+      stringsAsFactors = FALSE
+    )
+    result <- .build_consumer_map(metadata)
+    expect_type(result, "list")
+  })
+})
+
+describe(".find_upstream_deps", {
+  it("finds half.life via transitive chain: half.life -> lambda.z -> aucinf.obs", {
+    metadata <- data.frame(
+      PKNCA = c("half.life", "lambda.z", "aucinf.obs", "cmax", "tmax"),
+      Depends = c("tmax, tlast", "half.life", "lambda.z", NA, NA),
+      stringsAsFactors = FALSE
+    )
+    consumer_map <- .build_consumer_map(metadata)
+    start_set <- c("aucinf.obs")
+    result <- .find_upstream_deps(start_set, consumer_map)
+    expect_true("aucinf.obs" %in% result)
+    expect_true("lambda.z" %in% result)
+    expect_true("half.life" %in% result)
+    expect_false("cmax" %in% result)
+    expect_false("tmax" %in% result)
+  })
+
+  it("excludes observational leaf params (cmax, tmax, tlast) regardless of depth", {
+    metadata <- data.frame(
+      PKNCA = c("half.life", "lambda.z", "aucinf.obs"),
+      Depends = c("tmax", "half.life", "lambda.z"),
+      stringsAsFactors = FALSE
+    )
+    consumer_map <- .build_consumer_map(metadata)
+    result <- .find_upstream_deps(c("aucinf.obs"), consumer_map)
+    expect_true("lambda.z" %in% result)
+    expect_true("half.life" %in% result)
+    # tmax is in default obs_params → excluded regardless of depth
+    expect_false("tmax" %in% result)
+  })
+})
+
+describe("rm_impute_obs_params integration", {
+  it("retains imputation for half.life when aucinf.obs is requested (#1057)", {
+    data <- FIXTURE_PKNCA_DATA
+    # Mark all intervals as having imputation
+    if (!"impute" %in% names(data$intervals)) {
+      data$intervals$impute <- "blq"
+    } else {
+      data$intervals$impute[is.na(data$intervals$impute)] <- "blq"
+    }
+    # Request only aucinf.obs and half.life
+    for (col in names(data$intervals)) {
+      if (col %in% c("start", "end", "impute", "type_interval")) next
+      data$intervals[[col]] <- col %in% c("aucinf.obs", "half.life")
+    }
+    result <- rm_impute_obs_params(data, metadata_nca_parameters)
+    # half.life should retain imputation (it feeds aucinf.obs via lambda.z)
+    expect_true("blq" %in% result$intervals$impute)
+  })
+})