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: 0.1.0.9184
+Version: 0.1.0.9187
 Authors@R: c(
     person("Ercan", "Suekuer", email = "ercan.suekuer@roche.com", role = "aut",
            comment = c(ORCID = "0009-0001-1626-1526")),

NEWS.md

@@ -25,6 +25,9 @@
 * 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, and also walks downstream so AUC consumers (e.g., `vss.obs`, `vz.obs`) keep the same imputed data (#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 +330,26 @@ 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)
+
+  # Walk downstream too: AUC consumers (e.g., vss.obs = f(cl.obs, mrt.obs),
+  # vz.obs = f(cl.obs, lambda.z)) compute AUC-derived quantities internally,
+  # so they must use the same imputed data for consistency (#1057).
+  dependency_map <- .build_dependency_map(metadata_nca_parameters)
+  needs_impute <- .find_downstream_consumers(needs_impute, dependency_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 +383,95 @@ 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$Depends[i]
+    if (is.na(dep_str) || dep_str == "") next
+    dep_list <- trimws(strsplit(dep_str, ",")[[1]])
+    for (d in dep_list) {
+      rev[[d]] <- unique(c(rev[[d]], pkg_name))
+    }
+  }
+  rev
+}
+
+#' Build a dependency map from the Depends column.
+#' Mirror image of `.build_consumer_map()`: for each parameter A, returns the
+#' parameters A depends on.
+#' e.g., vss.obs lists "cl.obs, mrt.obs" in Depends, so
+#' dependency_map$vss.obs = c("cl.obs", "mrt.obs").
+#' @noRd
+.build_dependency_map <- function(metadata) {
+  fwd <- 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
+    fwd[[pkg_name]] <- trimws(strsplit(dep_str, ",")[[1]])
+  }
+  fwd
+}
+
+#' Check if a parameter is a new upstream consumer of the current chain.
+#' @noRd
+.is_new_upstream_consumer <- function(pkg, consumer_map, needs, obs_params) {
+  !pkg %in% needs && !pkg %in% obs_params && any(consumer_map[[pkg]] %in% needs)
+}
+
+#' 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 (.is_new_upstream_consumer(pkg, consumer_map, needs, obs_params)) {
+        newly_found <- c(newly_found, pkg)
+      }
+    }
+    if (length(newly_found) == 0) break
+    needs <- c(needs, newly_found)
+  }
+  needs
+}
+
+#' Find all downstream consumers transitively from `start_set`.
+#' Params whose Depends list intersects the current chain consume AUC-derived
+#' quantities (e.g., vss.obs consumes cl.obs and mrt.obs) and therefore compute
+#' AUC internally — they must use the same BLQ-imputed data (#1057).
+#'
+#' The fixpoint walk is direction-agnostic: `.find_upstream_deps()` follows
+#' "who consumes X" edges, so passing the reverse (dependency) map from
+#' `.build_dependency_map()` walks the graph downstream. Purely observational
+#' leaf params (cmax, tmax, tlast) are excluded as in the upstream walk.
+#'
+#' @param start_set Character vector of starting PKNCA parameter names.
+#' @param dependency_map Named list from `.build_dependency_map()`.
+#' @param obs_params Character vector of observational params to exclude.
+#' @param max_iter Maximum iterations to guard against infinite loops from
+#'   circular dependencies.
+#' @noRd
+.find_downstream_consumers <- function(start_set, dependency_map,
+                                       obs_params = c("cmax", "tmax", "tlast"),
+                                       max_iter = 50L) {
+  .find_upstream_deps(start_set, dependency_map, obs_params, max_iter)
+}

inst/WORDLIST

@@ -7,7 +7,9 @@ AEFRLT
 AFRLT
 ARRLT
 ATPTREF
+AUCIFO
 AUCIFOD
+AUCIFP
 AUCINT
 AUCLSTD
 AUCPEO
@@ -58,6 +60,7 @@ INTRAVASCULAR
 Kezia
 Kobana
 LAMZ
+LAMZHL
 LAMZLL
 LAMZNPT
 LAMZSPN

tests/testthat/data/test-blq-ADNCA.csv

@@ -0,0 +1,13 @@
+"STUDYID","USUBJID","PCSPEC","PARAM","DOSETRT","DOSEA","DOSEU","ROUTE","ADOSEDUR","AVAL","AVALU","AFRLT","NFRLT","ARRLT","NRRLT","RRLTU","ATPTREF","TRT01A","METABFL"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,100,"ng/mL",0.5,0.5,0.5,0.5,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,80,"ng/mL",1,1,1,1,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,65,"ng/mL",1.5,1.5,1.5,1.5,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,52,"ng/mL",2,2,2,2,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,35,"ng/mL",3,3,3,3,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,25,"ng/mL",4,4,4,4,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,12,"ng/mL",6,6,6,6,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,6,"ng/mL",8,8,8,8,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,2.5,"ng/mL",12,12,12,12,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,0,"ng/mL",24,24,24,24,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,0,"ng/mL",36,36,36,36,"hr","DOSE 1","DrugA 10 mg","N"
+"S1","BLQ-TEST-01","SERUM","DrugA","DrugA",10,"mg","INTRAVENOUS BOLUS",0,0,"ng/mL",48,48,48,48,"hr","DOSE 1","DrugA 10 mg","N"