diff --git a/R/clean_DIANN.R b/R/clean_DIANN.R
index 38cea27b..0ed98b05 100644
--- a/R/clean_DIANN.R
+++ b/R/clean_DIANN.R
@@ -221,32 +221,24 @@
 
 #' Assign IsotopeLabelType for DIANN protein turnover workflows.
 #'
-#' Dispatches to one of two classification paths depending on \code{has_channel}:
+#' Two paths, chosen by \code{has_channel}.  With a \code{Channel} column,
+#' \code{Channel} maps straight to \code{IsotopeLabelType} (\code{"H"} and
+#' \code{"L"} pass through, anything else becomes \code{NA}) and is then
+#' dropped; \code{labeledAminoAcids} acts only as the opt-in flag, and
+#' sequences are not inspected.
 #'
-#' \strong{Channel-based path} (\code{has_channel = TRUE}): \code{Channel}
-#' values are mapped directly to \code{IsotopeLabelType} (\code{"H"} →
-#' \code{"H"}, \code{"L"} → \code{"L"}, anything else → \code{NA}), and the
-#' \code{Channel} column is then dropped.  \code{labeledAminoAcids} acts solely
-#' as the opt-in flag that enables this path; the amino acid codes are
-#' \strong{not} used to validate or filter \code{ModifiedSequence}.
-#'
-#' \strong{ModifiedSequence-parsing path} (\code{has_channel = FALSE}):
-#' \code{PeptideSequence} (the retained \code{ModifiedSequence}) is scanned
-#' for isotope-labeled amino acids, which appear in parentheses immediately
-#' after the labeled residue, in the form \code{(<label>-<aminoAcid>-H)} for
-#' the heavy label or \code{(<label>-<aminoAcid>-L)} for the light label.
-#' For example, \code{K(label-K-H)} marks a heavy-labeled lysine (\code{K}).
-#' Here \code{<aminoAcid>} is one of the single-letter codes in
-#' \code{labeledAminoAcids}, and \code{<label>} is the label name (e.g.
-#' \code{SILAC} or \code{label}). Sequences with no such parenthetical
-#' tags are assigned \code{IsotopeLabelType = NA}. Once classified, the
-#' parenthetical annotation is stripped out of \code{PeptideSequence},
-#' leaving the plain amino acid sequence.
+#' Without one, the label is read from \code{PeptideSequence} (the retained
+#' \code{ModifiedSequence}), where it follows the labeled residue in
+#' parentheses as \code{(<label>-<aminoAcid>-H)} or
+#' \code{(<label>-<aminoAcid>-L)}; \code{K(label-K-H)} is a heavy-labeled
+#' lysine.  Untagged sequences become \code{NA}, the tags are then stripped
+#' from \code{PeptideSequence}, and peptides with two or more labelable
+#' residues (counted across all of \code{labeledAminoAcids}) are dropped, since
+#' partial labeling breaks the two-state turnover model.
 #'
 #' @param dn_input \code{data.table} after column renaming.
 #' @param labeledAminoAcids Character vector of single-letter amino acid codes
-#'   (e.g. \code{c("K")} or \code{c("K", "R")}), or \code{NULL} to skip
-#'   protein-turnover mode entirely (backwards-compatible default).
+#'   (e.g. \code{c("K", "R")}), or \code{NULL} to skip protein-turnover mode.
 #' @param has_channel Logical; \code{TRUE} when the raw input contained a
 #'   \code{Channel} column that was retained through
 #'   \code{.cleanDIANNSelectRequiredColumns}.
@@ -274,6 +266,8 @@
         light_regex <- paste0("\\([^-]+-(?:", aa_pattern, ")-L\\)")
         strip_regex <- paste0("\\([^-]+-(?:", aa_pattern, ")-[HL]\\)")
 
+        dn_input <- .filterMultiplyLabeledPeptides(dn_input, aa_pattern,
+                                                   "\\([^)]*\\)")
         dn_input <- .classifyIsotopeLabelType(dn_input, heavy_regex, light_regex)
         dn_input[, PeptideSequence := gsub(strip_regex, "", PeptideSequence, perl = TRUE)]
     }
diff --git a/R/clean_Spectronaut.R b/R/clean_Spectronaut.R
index 48b4287d..4c3a4ee6 100644
--- a/R/clean_Spectronaut.R
+++ b/R/clean_Spectronaut.R
@@ -171,27 +171,23 @@
 }
 
 
-#' Assign IsotopeLabelType based on heavy label detection.
+#' Assign IsotopeLabelType from Spectronaut heavy label tags.
 #'
-#' When \code{heavyLabel} is provided, each row is classified as heavy
-#' (\code{"H"}), light (\code{"L"}), or unlabeled (\code{NA}) by inspecting
-#' the labeled sequence column for the presence of the label tag.
+#' In Spectronaut turnover reports a labeled residue carries its label in
+#' brackets, e.g. \code{_PEPTIDEK[Lys6]_}.  Each \code{heavyLabels} entry pairs
+#' the residue with the label (\code{"K[Lys6]"}), so the bare residue marks a
+#' peptide that could have been labeled: with \code{"K[Lys6]"},
+#' \code{PEPTIDEK[Lys6]} is heavy, \code{PEPTIDEK} is light, and
+#' \code{PEPTIDEZ} is \code{NA}.  Peptides with two or more labelable residues
+#' (counted across all labels) are dropped, since partial labeling breaks the
+#' two-state turnover model.
 #'
-#' In Spectronaut protein turnover reports, heavy peptides appear in
-#' \code{FG.LabeledSequence} with a bracketed modification, e.g.
-#' \code{_PEPTIDEK[Lys6]_}.  Any sequence that contains
-#' \code{[<heavyLabel>]} is classified as heavy; all others are light.
-#' Sequences that do not have amino acids that can carry the label
-#' are classified as \code{NA}.  For example, if \code{heavyLabels} is 
-#' \code{"Lys6"}, then \code{PEPTIDEZ} is classified as NA since it
-#' has no lysine residues that could be labeled.
-#' When \code{heavyLabel} is \code{NULL} the column is left untouched so
-#' that the downstream \code{columns_to_fill} default of \code{"L"} applies,
-#' preserving backwards compatibility.
+#' With \code{NULL} the column is left untouched so the downstream
+#' \code{columns_to_fill} default of \code{"L"} applies.
 #'
 #' @param spec_input `data.table` after column renaming.
-#' @param heavyLabels Character scalar heavy label name (e.g. \code{"Lys6"}),
-#'   or \code{NULL}.
+#' @param heavyLabels Character vector of \code{<residue>[<label>]} pairs
+#'   (e.g. \code{"K[Lys6]"}), or \code{NULL}.
 #' @return `data.table` with \code{IsotopeLabelType} column added or updated.
 #' @keywords internal
 #' @noRd
@@ -208,6 +204,8 @@
         collapse = "|"
     )
 
+    spec_input = .filterMultiplyLabeledPeptides(spec_input, labeled_aa_regex,
+                                                "\\[[^\\]]*\\]")
     spec_input = .classifyIsotopeLabelType(spec_input, heavy_regex,
                                             labeled_aa_regex = labeled_aa_regex)
 
diff --git a/R/converters_DIANNtoMSstatsFormat.R b/R/converters_DIANNtoMSstatsFormat.R
index baddebbb..64cb90d8 100644
--- a/R/converters_DIANNtoMSstatsFormat.R
+++ b/R/converters_DIANNtoMSstatsFormat.R
@@ -19,37 +19,28 @@
 #' @param removeFewMeasurements should proteins with few measurements be removed
 #' @param removeOxidationMpeptides should peptides with oxidation be removed
 #' @param removeProtein_with1Feature should proteins with a single feature be removed
-#' @param labeledAminoAcids Character vector of single-letter amino acid codes
-#' that carry the SILAC label in protein turnover experiments, e.g.
-#' \code{c("K")} or \code{c("K", "R")}.  Supplying this vector opts in to
-#' protein-turnover mode; the exact amino acids determine behaviour only in the
-#' \code{ModifiedSequence}-parsing path described below.
+#' @param labeledAminoAcids Character vector of single-letter codes for the
+#' amino acids carrying the SILAC label, e.g. \code{c("K")} or
+#' \code{c("K", "R")}.  Supplying it turns on protein-turnover mode.  Defaults
+#' to \code{NULL}, which marks every peptide
+#' \code{IsotopeLabelType = "Light"}.  How the label is read depends on the
+#' export:
 #'
-#' \strong{Channel-based path} (DIA-NN 2.x exports that include a
-#' \code{Channel} column): when \code{labeledAminoAcids} is non-\code{NULL}
-#' \emph{and} the input contains a \code{Channel} column, \code{Channel} values
-#' are mapped directly to \code{IsotopeLabelType} (\code{"H"} → \code{"H"},
-#' \code{"L"} → \code{"L"}, anything else → \code{NA}).  The amino acid codes
-#' in \code{labeledAminoAcids} are \strong{not} used to validate or filter
-#' \code{ModifiedSequence} in this path.
+#' \strong{DIA-NN 2.x, with a \code{Channel} column.}  \code{Channel} gives the
+#' label directly: \code{"H"} and \code{"L"} pass through, anything else
+#' becomes \code{NA}.  The amino acid codes only switch the mode on here;
+#' sequences are not inspected.
 #'
-#' \strong{ModifiedSequence-parsing path} (DIA-NN 1.x exports without a
-#' \code{Channel} column): when \code{labeledAminoAcids} is non-\code{NULL}
-#' and no \code{Channel} column is present, each \code{ModifiedSequence} is 
-#' scanned for isotope-labeled amino acids, which appear in parentheses 
-#' immediately after the labeled residue, 
-#' in the form \code{(<label>-<aminoAcid>-H)} for
-#' the heavy label or \code{(<label>-<aminoAcid>-L)} for the light label.
-#' For example, \code{K(label-K-H)} marks a heavy-labeled lysine (\code{K}).
-#' Here \code{<aminoAcid>} is one of the single-letter codes in
-#' \code{labeledAminoAcids}, and \code{<label>} is the label name (e.g.
-#' \code{SILAC} or \code{label}). Sequences with no such parenthetical
-#' tags are assigned \code{IsotopeLabelType = NA}. Once classified, the
-#' parenthetical annotation is stripped out of \code{PeptideSequence},
-#' leaving the plain amino acid sequence.
-#'
-#' When \code{NULL} (default), protein-turnover mode is disabled and all
-#' peptides receive \code{IsotopeLabelType = "Light"}.
+#' \strong{DIA-NN 1.x, without a \code{Channel} column.}  The label is read
+#' from \code{ModifiedSequence}, where it follows the labeled residue in
+#' parentheses as \code{(<label>-<aminoAcid>-H)} or
+#' \code{(<label>-<aminoAcid>-L)}; for example \code{K(label-K-H)} is a
+#' heavy-labeled lysine.  \code{<label>} is the label name DIA-NN reported
+#' (e.g. \code{SILAC}).  Untagged sequences become \code{NA}, and the tags are
+#' stripped from \code{PeptideSequence} afterwards.  A peptide with two or more
+#' labelable residues (counted across all of \code{labeledAminoAcids}) may be
+#' only partially labeled, which the two-state turnover model cannot represent,
+#' so it is dropped and the number removed is reported.
 #' @param quantificationColumn Use 'FragmentQuantCorrected'(default) column for quantified intensities for DIANN 1.8.x.
 #' Use 'FragmentQuantRaw' for quantified intensities for DIANN 1.9.x.
 #' Use 'auto' for quantified intensities for DIANN 2.x where each fragment intensity is a separate column, e.g. Fr0Quantity.
diff --git a/R/converters_SpectronauttoMSstatsFormat.R b/R/converters_SpectronauttoMSstatsFormat.R
index b929b0e3..a45b76eb 100644
--- a/R/converters_SpectronauttoMSstatsFormat.R
+++ b/R/converters_SpectronauttoMSstatsFormat.R
@@ -12,17 +12,23 @@
 #' @param peptideSequenceColumn Name of the Spectronaut column that contains the
 #' peptide sequence.  Defaults to \code{"EG.ModifiedSequence"}. The value is 
 #' standardized internally (dots and spaces removed) before column lookup.
-#' @param heavyLabels Character list identifying the heavy isotope labels as it
-#'   appears inside square brackets in the peptide sequence column, e.g.
-#'   \code{c("Lys6")} matches peptides containing \code{[Lys6]}.  
-#'   \code{c("Lys6", "Arg10")} matches peptides containing either \code{[Lys6]} or \code{[Arg10]}.
-#'   Supports any novel label name reported by Spectronaut (e.g. \code{"Leu6"},
-#'   \code{"Phe10"}).  When provided, peptides are
-#'   classified as heavy (\code{IsotopeLabelType = "H"}), light
-#'   (\code{IsotopeLabelType = "L"}), or unlabeled
-#'   (\code{IsotopeLabelType = NA}) based on its labeled sequence.  When
-#'   \code{NULL} (default) all peptides receive \code{IsotopeLabelType = "L"}.
-#'   Useful for protein turnover experiments.
+#' @param heavyLabels Character vector naming each labeled residue and its
+#'   heavy label as they appear in the peptide sequence column: the
+#'   single-letter amino acid code, then the label in square brackets.  For
+#'   example \code{"K[Lys6]"}, or \code{c("K[Lys6]", "R[Arg10]")} for a
+#'   double-label experiment.  Any label name Spectronaut reports is accepted.
+#'   The residue letter is required, because it is what tells MSstats which
+#'   peptides could have carried a label.
+#'
+#'   Peptides carrying the label are marked heavy
+#'   (\code{IsotopeLabelType = "H"}), those that could carry it but do not are
+#'   light (\code{"L"}), and those with no labelable residue are \code{NA}.  A
+#'   peptide with two or more labelable residues (counted across all labels
+#'   supplied) may be only partially labeled, which the two-state turnover
+#'   model cannot represent, so it is dropped and the number removed is
+#'   reported.
+#'
+#'   Defaults to \code{NULL}: turnover mode off, every peptide marked light.
 #' @param excludedFromQuantificationFilter Remove rows with F.ExcludedFromQuantification=TRUE Default is TRUE.
 #' @param filter_with_Qvalue FALSE(default) will not perform any filtering. TRUE will filter out the intensities that have greater than qvalue_cutoff in EG.Qvalue column. Those intensities will be replaced with zero and will be considered as censored missing values for imputation purpose.
 #' @param qvalue_cutoff Cutoff for EG.Qvalue. default is 0.01.
diff --git a/R/utils_clean_features.R b/R/utils_clean_features.R
index c1c7094a..2928909c 100644
--- a/R/utils_clean_features.R
+++ b/R/utils_clean_features.R
@@ -309,22 +309,63 @@
 }
 
 
+#' Count regex matches per element, scoring no-match and \code{NA} as 0.
+#' @param x Character vector to search.
+#' @param pattern Perl-compatible regex.
+#' @return Integer vector of match counts, the same length as \code{x}.
+#' @keywords internal
+#' @noRd
+.countRegexMatches = function(x, pattern) {
+    lengths(regmatches(x, gregexpr(pattern, x, perl = TRUE)))
+}
+
+
+#' Drop peptides carrying more than one labelable residue.
+#'
+#' Such peptides can be partially labeled, which the two-state turnover model
+#' cannot represent, so heavy and light rows are dropped together to keep the
+#' light/heavy ratio unbiased.  The number of peptides removed is logged, since
+#' the exclusion is otherwise invisible to the user.
+#'
+#' @param dt \code{data.table} with a \code{PeptideSequence} column.
+#' @param residue_regex Perl-compatible regex matching one labelable residue.
+#' @param strip_regex Perl-compatible regex matching label and modification
+#'   annotations, removed before counting so that residue letters inside an
+#'   annotation are not counted.
+#' @return \code{dt} with multiply labeled rows removed.
+#' @keywords internal
+#' @noRd
+.filterMultiplyLabeledPeptides = function(dt, residue_regex, strip_regex) {
+    stripped = gsub(strip_regex, "", dt[["PeptideSequence"]], perl = TRUE)
+    n_labelable = .countRegexMatches(stripped, residue_regex)
+    is_multiply_labeled = n_labelable >= 2L
+
+    if (any(is_multiply_labeled)) {
+        # Count distinct peptides on the stripped sequence, so that the heavy
+        # and light forms of one peptide are not reported as two.
+        msg = paste("**", data.table::uniqueN(stripped[is_multiply_labeled]),
+                    "peptide(s) with more than one labelable residue were",
+                    paste0("removed (", sum(is_multiply_labeled), " row(s))."),
+                    "Turnover analysis is currently limited to peptides with",
+                    "exactly one labelable residue.")
+        getOption("MSstatsLog")("INFO", msg)
+        getOption("MSstatsMsg")("INFO", msg)
+    }
+
+    dt[!is_multiply_labeled, ]
+}
+
+
 #' Classify IsotopeLabelType from peptide sequence patterns.
 #'
-#' Shared core logic for protein turnover workflows in both Spectronaut and
-#' DIANN converters.  Each peptide is classified as heavy (\code{"H"}), light
-#' (\code{"L"}), or unlabeled (\code{NA}) by matching regex patterns against
-#' the \code{PeptideSequence} column.
+#' Shared by the Spectronaut and DIANN turnover workflows.  Each peptide is
+#' classified as heavy (\code{"H"}), light (\code{"L"}), or unlabeled
+#' (\code{NA}) by matching regexes against \code{PeptideSequence}.
 #'
-#' Two modes are supported:
-#' \describe{
-#'   \item{Spectronaut mode}{Pass \code{labeled_aa_regex}.  The sequence is
-#'     first stripped of all bracket modifications; light is inferred when the
-#'     bare labeled amino acid is present but the heavy bracket form is absent.}
-#'   \item{DIANN mode}{Pass \code{light_regex}.  Both heavy and light patterns
-#'     are matched directly against the modified sequence; absence of either
-#'     yields \code{NA}.}
-#' }
+#' Spectronaut mode (\code{labeled_aa_regex}) strips bracket modifications
+#' first, then infers light from a bare labelable residue with no heavy tag.
+#' DIANN mode (\code{light_regex}) matches heavy and light patterns directly
+#' against the modified sequence.
 #'
 #' @param dt \code{data.table} with a \code{PeptideSequence} column.
 #' @param heavy_regex Perl-compatible regex matching heavy-labeled sequences.
diff --git a/inst/tinytest/test_clean_DIANN.R b/inst/tinytest/test_clean_DIANN.R
index 64c63f2e..0d9ba80f 100644
--- a/inst/tinytest/test_clean_DIANN.R
+++ b/inst/tinytest/test_clean_DIANN.R
@@ -99,3 +99,54 @@ expect_equal(result_multi_aa$IsotopeLabelType,
              c("H", "H", "L", "L", NA_character_))
 expect_equal(sort(unique(result_multi_aa$PeptideSequence)),
              c("PEPTIDEAC", "PEPTIDEK", "PEPTIDER"))
+
+# Multiply labeled peptides (2+ labelable residues) are filtered out
+dt_multi_label = data.table::data.table(
+    PeptideSequence = c(
+        "PEPTIDEK(SILAC-K-H)",                    # 1 K, heavy -> kept
+        "PEPTIDEK(SILAC-K-L)",                    # 1 K, light -> kept
+        "PEPK(SILAC-K-H)TIDEK(SILAC-K-H)",        # 2 K, heavy -> dropped
+        "PEPK(SILAC-K-H)TIDEK(SILAC-K-L)",        # 2 K, partial -> dropped
+        "PEPK(SILAC-K-L)TIDEK(SILAC-K-L)",        # 2 K, light -> dropped
+        "PEPTIDEAC"                               # 0 K -> kept as NA
+    )
+)
+result_multi_label = MSstatsConvert:::.assignDIANNIsotopeLabelType(
+    dt_multi_label, labeledAminoAcids = c("K"), has_channel = FALSE
+)
+expect_equal(result_multi_label$PeptideSequence,
+             c("PEPTIDEK", "PEPTIDEK", "PEPTIDEAC"))
+expect_equal(result_multi_label$IsotopeLabelType, c("H", "L", NA_character_))
+
+# Count is taken across all labeled amino acids combined
+dt_kr = data.table::data.table(
+    PeptideSequence = c("PEPTIDEK(SILAC-K-H)",
+                        "PEPK(SILAC-K-H)TIDER(SILAC-R-H)",  # 1 K + 1 R -> dropped
+                        "PEPTIDER(SILAC-R-L)")
+)
+result_kr = MSstatsConvert:::.assignDIANNIsotopeLabelType(
+    dt_kr, labeledAminoAcids = c("K", "R"), has_channel = FALSE
+)
+expect_equal(result_kr$PeptideSequence, c("PEPTIDEK", "PEPTIDER"))
+expect_equal(result_kr$IsotopeLabelType, c("H", "L"))
+
+# Residue letters inside an unrelated modification are not counted
+dt_mod = data.table::data.table(
+    PeptideSequence = c("PEPTS(Kmod)IDEK(SILAC-K-H)")
+)
+result_mod = MSstatsConvert:::.assignDIANNIsotopeLabelType(
+    dt_mod, labeledAminoAcids = c("K"), has_channel = FALSE
+)
+expect_equal(nrow(result_mod), 1L)
+expect_equal(result_mod$IsotopeLabelType, "H")
+
+# The Channel path is exempt: labeling is not inferred from sequence content
+dt_channel_multi = data.table::data.table(
+    PeptideSequence = c("PEPKTIDEK", "PEPKTIDEK", "PEPTIDEK"),
+    Channel = c("H", "L", "H")
+)
+result_channel_multi = MSstatsConvert:::.assignDIANNIsotopeLabelType(
+    dt_channel_multi, labeledAminoAcids = c("K"), has_channel = TRUE
+)
+expect_equal(nrow(result_channel_multi), 3L)
+expect_equal(result_channel_multi$IsotopeLabelType, c("H", "L", "H"))
diff --git a/inst/tinytest/test_clean_Spectronaut.R b/inst/tinytest/test_clean_Spectronaut.R
index a4a581a7..e1df964c 100644
--- a/inst/tinytest/test_clean_Spectronaut.R
+++ b/inst/tinytest/test_clean_Spectronaut.R
@@ -73,3 +73,37 @@ expect_equal(result$PeptideSequence,
 dt = make_spec_input(c("_PEPTIDEK_", "_PEPTIDER_"))
 result = MSstatsConvert:::.assignSpectronautIsotopeLabelType(dt, heavyLabels = NULL)
 expect_equal(result, dt)
+
+# Multiply labeled peptides (2+ labelable residues) are filtered out
+dt = make_spec_input(c(
+    "_PEPTIDEK[Lys6]_",         # 1 K, heavy -> kept
+    "_PEPTIDEK_",               # 1 K, light -> kept
+    "_PEPK[Lys6]TIDEK[Lys6]_",  # 2 K, fully heavy -> dropped
+    "_PEPK[Lys6]TIDEK_",        # 2 K, partially labeled -> dropped
+    "_PEPKTIDEK_",              # 2 K, fully light -> dropped
+    "_ACDEGFHI_"                # 0 K -> kept as NA
+))
+result = MSstatsConvert:::.assignSpectronautIsotopeLabelType(
+    dt, heavyLabels = "K[Lys6]")
+expect_equal(result$PeptideSequence,
+             c("_PEPTIDEK_", "_PEPTIDEK_", "_ACDEGFHI_"))
+expect_equal(result$IsotopeLabelType, c("H", "L", NA_character_))
+
+# Count is taken across all labelable residues combined: one K plus one R is
+# doubly labelable when both labels are specified
+dt = make_spec_input(c(
+    "_PEPTIDEK_",   # 1 labelable -> kept
+    "_PEPTIDEKR_",  # 1 K + 1 R -> dropped
+    "_PEPTIDER_"    # 1 labelable -> kept
+))
+result = MSstatsConvert:::.assignSpectronautIsotopeLabelType(
+    dt, heavyLabels = c("K[Lys6]", "R[Arg10]"))
+expect_equal(result$PeptideSequence, c("_PEPTIDEK_", "_PEPTIDER_"))
+expect_equal(result$IsotopeLabelType, c("L", "L"))
+
+# Residue letters inside an unrelated modification tag are not counted
+dt = make_spec_input(c("_S[Kmodification]PEPTIDEK_"))
+result = MSstatsConvert:::.assignSpectronautIsotopeLabelType(
+    dt, heavyLabels = "K[Lys6]")
+expect_equal(nrow(result), 1L)
+expect_equal(result$IsotopeLabelType, "L")
diff --git a/inst/tinytest/test_utils_clean_features.R b/inst/tinytest/test_utils_clean_features.R
index aec83c3f..4070faa8 100644
--- a/inst/tinytest/test_utils_clean_features.R
+++ b/inst/tinytest/test_utils_clean_features.R
@@ -216,6 +216,32 @@ result_multi = MSstatsConvert:::.classifyIsotopeLabelType(
 )
 expect_equal(result_multi$IsotopeLabelType, c("H", "H", "L", "L", NA_character_))
 
+# Test .filterMultiplyLabeledPeptides ----
+# Heavy and light forms of one peptide are reported as a single peptide, and
+# the row count is reported alongside it
+dt_report = data.table::data.table(PeptideSequence = c(
+    "_PEPKTIDEK[Lys6]_", "_PEPKTIDEK[Lys6]_", "_PEPKTIDEK_",  # 1 peptide, 3 rows
+    "_PEPTIDEK_"                                              # kept
+))
+# The appender writes with cat(), so the console output is captured directly
+log_report = capture.output(
+    kept <- MSstatsConvert:::.filterMultiplyLabeledPeptides(
+        dt_report, "K", "\\[.*?\\]")
+)
+expect_true(any(grepl(
+    "1 peptide(s) with more than one labelable residue were removed (3 row(s))",
+    log_report, fixed = TRUE)))
+expect_equal(nrow(kept), 1L)
+
+# Nothing to remove means nothing is reported
+dt_quiet = data.table::data.table(PeptideSequence = c("_PEPTIDEK_", "_ACDEG_"))
+log_quiet = capture.output(
+    all_kept <- MSstatsConvert:::.filterMultiplyLabeledPeptides(
+        dt_quiet, "K", "\\[.*?\\]")
+)
+expect_false(any(grepl("labelable residue", log_quiet, fixed = TRUE)))
+expect_equal(nrow(all_kept), 2L)
+
 # Utility function ----
 expect_equal(MSstatsConvert:::.combine(c("A", "B"), c("A", "B")),
              c("A_A", "B_B"))
diff --git a/man/DIANNtoMSstatsFormat.Rd b/man/DIANNtoMSstatsFormat.Rd
index 5fe18013..521877e7 100644
--- a/man/DIANNtoMSstatsFormat.Rd
+++ b/man/DIANNtoMSstatsFormat.Rd
@@ -62,37 +62,28 @@ the library created after the first MBR pass. Default is 0.01.}
 
 \item{MBR}{True if analysis was done with match between runs}
 
-\item{labeledAminoAcids}{Character vector of single-letter amino acid codes
-that carry the SILAC label in protein turnover experiments, e.g.
-\code{c("K")} or \code{c("K", "R")}.  Supplying this vector opts in to
-protein-turnover mode; the exact amino acids determine behaviour only in the
-\code{ModifiedSequence}-parsing path described below.
-
-\strong{Channel-based path} (DIA-NN 2.x exports that include a
-\code{Channel} column): when \code{labeledAminoAcids} is non-\code{NULL}
-\emph{and} the input contains a \code{Channel} column, \code{Channel} values
-are mapped directly to \code{IsotopeLabelType} (\code{"H"} → \code{"H"},
-\code{"L"} → \code{"L"}, anything else → \code{NA}).  The amino acid codes
-in \code{labeledAminoAcids} are \strong{not} used to validate or filter
-\code{ModifiedSequence} in this path.
-
-\strong{ModifiedSequence-parsing path} (DIA-NN 1.x exports without a
-\code{Channel} column): when \code{labeledAminoAcids} is non-\code{NULL}
-and no \code{Channel} column is present, each \code{ModifiedSequence} is
-scanned for isotope-labeled amino acids, which appear in parentheses
-immediately after the labeled residue,
-in the form \code{(<label>-<aminoAcid>-H)} for
-the heavy label or \code{(<label>-<aminoAcid>-L)} for the light label.
-For example, \code{K(label-K-H)} marks a heavy-labeled lysine (\code{K}).
-Here \code{<aminoAcid>} is one of the single-letter codes in
-\code{labeledAminoAcids}, and \code{<label>} is the label name (e.g.
-\code{SILAC} or \code{label}). Sequences with no such parenthetical
-tags are assigned \code{IsotopeLabelType = NA}. Once classified, the
-parenthetical annotation is stripped out of \code{PeptideSequence},
-leaving the plain amino acid sequence.
-
-When \code{NULL} (default), protein-turnover mode is disabled and all
-peptides receive \code{IsotopeLabelType = "Light"}.}
+\item{labeledAminoAcids}{Character vector of single-letter codes for the
+amino acids carrying the SILAC label, e.g. \code{c("K")} or
+\code{c("K", "R")}.  Supplying it turns on protein-turnover mode.  Defaults
+to \code{NULL}, which marks every peptide
+\code{IsotopeLabelType = "Light"}.  How the label is read depends on the
+export:
+
+\strong{DIA-NN 2.x, with a \code{Channel} column.}  \code{Channel} gives the
+label directly: \code{"H"} and \code{"L"} pass through, anything else
+becomes \code{NA}.  The amino acid codes only switch the mode on here;
+sequences are not inspected.
+
+\strong{DIA-NN 1.x, without a \code{Channel} column.}  The label is read
+from \code{ModifiedSequence}, where it follows the labeled residue in
+parentheses as \code{(<label>-<aminoAcid>-H)} or
+\code{(<label>-<aminoAcid>-L)}; for example \code{K(label-K-H)} is a
+heavy-labeled lysine.  \code{<label>} is the label name DIA-NN reported
+(e.g. \code{SILAC}).  Untagged sequences become \code{NA}, and the tags are
+stripped from \code{PeptideSequence} afterwards.  A peptide with two or more
+labelable residues (counted across all of \code{labeledAminoAcids}) may be
+only partially labeled, which the two-state turnover model cannot represent,
+so it is dropped and the number removed is reported.}
 
 \item{quantificationColumn}{Use 'FragmentQuantCorrected'(default) column for quantified intensities for DIANN 1.8.x.
 Use 'FragmentQuantRaw' for quantified intensities for DIANN 1.9.x.
diff --git a/man/MSstatsClean.Rd b/man/MSstatsClean.Rd
index d8cccabe..8e154095 100644
--- a/man/MSstatsClean.Rd
+++ b/man/MSstatsClean.Rd
@@ -137,17 +137,23 @@ For protein turnover workflows the recommended default is
 peptide sequence.  Defaults to \code{"EG.ModifiedSequence"}. The value is
 standardized internally (dots and spaces removed) before column lookup.}
 
-\item{heavyLabels}{Character list identifying the heavy isotope labels as it
-appears inside square brackets in the peptide sequence column, e.g.
-\code{c("Lys6")} matches peptides containing \code{[Lys6]}.
-\code{c("Lys6", "Arg10")} matches peptides containing either \code{[Lys6]} or \code{[Arg10]}.
-Supports any novel label name reported by Spectronaut (e.g. \code{"Leu6"},
-\code{"Phe10"}).  When provided, peptides are
-classified as heavy (\code{IsotopeLabelType = "H"}), light
-(\code{IsotopeLabelType = "L"}), or unlabeled
-(\code{IsotopeLabelType = NA}) based on its labeled sequence.  When
-\code{NULL} (default) all peptides receive \code{IsotopeLabelType = "L"}.
-Useful for protein turnover experiments.}
+\item{heavyLabels}{Character vector naming each labeled residue and its
+heavy label as they appear in the peptide sequence column: the
+single-letter amino acid code, then the label in square brackets.  For
+example \code{"K[Lys6]"}, or \code{c("K[Lys6]", "R[Arg10]")} for a
+double-label experiment.  Any label name Spectronaut reports is accepted.
+The residue letter is required, because it is what tells MSstats which
+peptides could have carried a label.
+
+Peptides carrying the label are marked heavy
+(\code{IsotopeLabelType = "H"}), those that could carry it but do not are
+light (\code{"L"}), and those with no labelable residue are \code{NA}.  A
+peptide with two or more labelable residues (counted across all labels
+supplied) may be only partially labeled, which the two-state turnover
+model cannot represent, so it is dropped and the number removed is
+reported.
+
+Defaults to \code{NULL}: turnover mode off, every peptide marked light.}
 
 \item{peptide_id_col}{character name of a column that identifies peptides}
 
@@ -175,37 +181,28 @@ column, i.e. the global q-value for the protein group.  If MBR is true, the
 qvalue cutoff for the Lib.PG.Q.Value column, i.e. the protein group q-value for
 the library created after the first MBR pass. Default is 0.01.}
 
-\item{labeledAminoAcids}{Character vector of single-letter amino acid codes
-that carry the SILAC label in protein turnover experiments, e.g.
-\code{c("K")} or \code{c("K", "R")}.  Supplying this vector opts in to
-protein-turnover mode; the exact amino acids determine behaviour only in the
-\code{ModifiedSequence}-parsing path described below.
-
-\strong{Channel-based path} (DIA-NN 2.x exports that include a
-\code{Channel} column): when \code{labeledAminoAcids} is non-\code{NULL}
-\emph{and} the input contains a \code{Channel} column, \code{Channel} values
-are mapped directly to \code{IsotopeLabelType} (\code{"H"} → \code{"H"},
-\code{"L"} → \code{"L"}, anything else → \code{NA}).  The amino acid codes
-in \code{labeledAminoAcids} are \strong{not} used to validate or filter
-\code{ModifiedSequence} in this path.
-
-\strong{ModifiedSequence-parsing path} (DIA-NN 1.x exports without a
-\code{Channel} column): when \code{labeledAminoAcids} is non-\code{NULL}
-and no \code{Channel} column is present, each \code{ModifiedSequence} is
-scanned for isotope-labeled amino acids, which appear in parentheses
-immediately after the labeled residue,
-in the form \code{(<label>-<aminoAcid>-H)} for
-the heavy label or \code{(<label>-<aminoAcid>-L)} for the light label.
-For example, \code{K(label-K-H)} marks a heavy-labeled lysine (\code{K}).
-Here \code{<aminoAcid>} is one of the single-letter codes in
-\code{labeledAminoAcids}, and \code{<label>} is the label name (e.g.
-\code{SILAC} or \code{label}). Sequences with no such parenthetical
-tags are assigned \code{IsotopeLabelType = NA}. Once classified, the
-parenthetical annotation is stripped out of \code{PeptideSequence},
-leaving the plain amino acid sequence.
-
-When \code{NULL} (default), protein-turnover mode is disabled and all
-peptides receive \code{IsotopeLabelType = "Light"}.}
+\item{labeledAminoAcids}{Character vector of single-letter codes for the
+amino acids carrying the SILAC label, e.g. \code{c("K")} or
+\code{c("K", "R")}.  Supplying it turns on protein-turnover mode.  Defaults
+to \code{NULL}, which marks every peptide
+\code{IsotopeLabelType = "Light"}.  How the label is read depends on the
+export:
+
+\strong{DIA-NN 2.x, with a \code{Channel} column.}  \code{Channel} gives the
+label directly: \code{"H"} and \code{"L"} pass through, anything else
+becomes \code{NA}.  The amino acid codes only switch the mode on here;
+sequences are not inspected.
+
+\strong{DIA-NN 1.x, without a \code{Channel} column.}  The label is read
+from \code{ModifiedSequence}, where it follows the labeled residue in
+parentheses as \code{(<label>-<aminoAcid>-H)} or
+\code{(<label>-<aminoAcid>-L)}; for example \code{K(label-K-H)} is a
+heavy-labeled lysine.  \code{<label>} is the label name DIA-NN reported
+(e.g. \code{SILAC}).  Untagged sequences become \code{NA}, and the tags are
+stripped from \code{PeptideSequence} afterwards.  A peptide with two or more
+labelable residues (counted across all of \code{labeledAminoAcids}) may be
+only partially labeled, which the two-state turnover model cannot represent,
+so it is dropped and the number removed is reported.}
 
 \item{mzmine_annotations}{\code{data.frame} of MZMine spectral-library
 annotations with columns \code{id}, \code{compound_name}, \code{score}. Required;
diff --git a/man/SpectronauttoMSstatsFormat.Rd b/man/SpectronauttoMSstatsFormat.Rd
index 91caca28..2233318e 100644
--- a/man/SpectronauttoMSstatsFormat.Rd
+++ b/man/SpectronauttoMSstatsFormat.Rd
@@ -50,17 +50,23 @@ For protein turnover workflows the recommended default is
 peptide sequence.  Defaults to \code{"EG.ModifiedSequence"}. The value is
 standardized internally (dots and spaces removed) before column lookup.}
 
-\item{heavyLabels}{Character list identifying the heavy isotope labels as it
-appears inside square brackets in the peptide sequence column, e.g.
-\code{c("Lys6")} matches peptides containing \code{[Lys6]}.
-\code{c("Lys6", "Arg10")} matches peptides containing either \code{[Lys6]} or \code{[Arg10]}.
-Supports any novel label name reported by Spectronaut (e.g. \code{"Leu6"},
-\code{"Phe10"}).  When provided, peptides are
-classified as heavy (\code{IsotopeLabelType = "H"}), light
-(\code{IsotopeLabelType = "L"}), or unlabeled
-(\code{IsotopeLabelType = NA}) based on its labeled sequence.  When
-\code{NULL} (default) all peptides receive \code{IsotopeLabelType = "L"}.
-Useful for protein turnover experiments.}
+\item{heavyLabels}{Character vector naming each labeled residue and its
+heavy label as they appear in the peptide sequence column: the
+single-letter amino acid code, then the label in square brackets.  For
+example \code{"K[Lys6]"}, or \code{c("K[Lys6]", "R[Arg10]")} for a
+double-label experiment.  Any label name Spectronaut reports is accepted.
+The residue letter is required, because it is what tells MSstats which
+peptides could have carried a label.
+
+Peptides carrying the label are marked heavy
+(\code{IsotopeLabelType = "H"}), those that could carry it but do not are
+light (\code{"L"}), and those with no labelable residue are \code{NA}.  A
+peptide with two or more labelable residues (counted across all labels
+supplied) may be only partially labeled, which the two-state turnover
+model cannot represent, so it is dropped and the number removed is
+reported.
+
+Defaults to \code{NULL}: turnover mode off, every peptide marked light.}
 
 \item{excludedFromQuantificationFilter}{Remove rows with F.ExcludedFromQuantification=TRUE Default is TRUE.}
 
diff --git a/man/dot-cleanRawDIANN.Rd b/man/dot-cleanRawDIANN.Rd
index 92bcf043..b541d479 100644
--- a/man/dot-cleanRawDIANN.Rd
+++ b/man/dot-cleanRawDIANN.Rd
@@ -42,37 +42,28 @@ the library created after the first MBR pass. Default is 0.01.}
 
 \item{anomalyModelFeatures}{character vector of quality metric column names to be used as features in the anomaly detection model. List must not be empty if calculateAnomalyScores=TRUE.}
 
-\item{labeledAminoAcids}{Character vector of single-letter amino acid codes
-that carry the SILAC label in protein turnover experiments, e.g.
-\code{c("K")} or \code{c("K", "R")}.  Supplying this vector opts in to
-protein-turnover mode; the exact amino acids determine behaviour only in the
-\code{ModifiedSequence}-parsing path described below.
+\item{labeledAminoAcids}{Character vector of single-letter codes for the
+amino acids carrying the SILAC label, e.g. \code{c("K")} or
+\code{c("K", "R")}.  Supplying it turns on protein-turnover mode.  Defaults
+to \code{NULL}, which marks every peptide
+\code{IsotopeLabelType = "Light"}.  How the label is read depends on the
+export:
 
-\strong{Channel-based path} (DIA-NN 2.x exports that include a
-\code{Channel} column): when \code{labeledAminoAcids} is non-\code{NULL}
-\emph{and} the input contains a \code{Channel} column, \code{Channel} values
-are mapped directly to \code{IsotopeLabelType} (\code{"H"} → \code{"H"},
-\code{"L"} → \code{"L"}, anything else → \code{NA}).  The amino acid codes
-in \code{labeledAminoAcids} are \strong{not} used to validate or filter
-\code{ModifiedSequence} in this path.
+\strong{DIA-NN 2.x, with a \code{Channel} column.}  \code{Channel} gives the
+label directly: \code{"H"} and \code{"L"} pass through, anything else
+becomes \code{NA}.  The amino acid codes only switch the mode on here;
+sequences are not inspected.
 
-\strong{ModifiedSequence-parsing path} (DIA-NN 1.x exports without a
-\code{Channel} column): when \code{labeledAminoAcids} is non-\code{NULL}
-and no \code{Channel} column is present, each \code{ModifiedSequence} is
-scanned for isotope-labeled amino acids, which appear in parentheses
-immediately after the labeled residue,
-in the form \code{(<label>-<aminoAcid>-H)} for
-the heavy label or \code{(<label>-<aminoAcid>-L)} for the light label.
-For example, \code{K(label-K-H)} marks a heavy-labeled lysine (\code{K}).
-Here \code{<aminoAcid>} is one of the single-letter codes in
-\code{labeledAminoAcids}, and \code{<label>} is the label name (e.g.
-\code{SILAC} or \code{label}). Sequences with no such parenthetical
-tags are assigned \code{IsotopeLabelType = NA}. Once classified, the
-parenthetical annotation is stripped out of \code{PeptideSequence},
-leaving the plain amino acid sequence.
-
-When \code{NULL} (default), protein-turnover mode is disabled and all
-peptides receive \code{IsotopeLabelType = "Light"}.}
+\strong{DIA-NN 1.x, without a \code{Channel} column.}  The label is read
+from \code{ModifiedSequence}, where it follows the labeled residue in
+parentheses as \code{(<label>-<aminoAcid>-H)} or
+\code{(<label>-<aminoAcid>-L)}; for example \code{K(label-K-H)} is a
+heavy-labeled lysine.  \code{<label>} is the label name DIA-NN reported
+(e.g. \code{SILAC}).  Untagged sequences become \code{NA}, and the tags are
+stripped from \code{PeptideSequence} afterwards.  A peptide with two or more
+labelable residues (counted across all of \code{labeledAminoAcids}) may be
+only partially labeled, which the two-state turnover model cannot represent,
+so it is dropped and the number removed is reported.}
 }
 \value{
 data.table
diff --git a/man/dot-cleanRawSpectronaut.Rd b/man/dot-cleanRawSpectronaut.Rd
index edc51a3a..39028cf7 100644
--- a/man/dot-cleanRawSpectronaut.Rd
+++ b/man/dot-cleanRawSpectronaut.Rd
@@ -32,17 +32,23 @@ For protein turnover workflows the recommended default is
 peptide sequence.  Defaults to \code{"EG.ModifiedSequence"}. The value is
 standardized internally (dots and spaces removed) before column lookup.}
 
-\item{heavyLabels}{Character list identifying the heavy isotope labels as it
-appears inside square brackets in the peptide sequence column, e.g.
-\code{c("Lys6")} matches peptides containing \code{[Lys6]}.
-\code{c("Lys6", "Arg10")} matches peptides containing either \code{[Lys6]} or \code{[Arg10]}.
-Supports any novel label name reported by Spectronaut (e.g. \code{"Leu6"},
-\code{"Phe10"}).  When provided, peptides are
-classified as heavy (\code{IsotopeLabelType = "H"}), light
-(\code{IsotopeLabelType = "L"}), or unlabeled
-(\code{IsotopeLabelType = NA}) based on its labeled sequence.  When
-\code{NULL} (default) all peptides receive \code{IsotopeLabelType = "L"}.
-Useful for protein turnover experiments.}
+\item{heavyLabels}{Character vector naming each labeled residue and its
+heavy label as they appear in the peptide sequence column: the
+single-letter amino acid code, then the label in square brackets.  For
+example \code{"K[Lys6]"}, or \code{c("K[Lys6]", "R[Arg10]")} for a
+double-label experiment.  Any label name Spectronaut reports is accepted.
+The residue letter is required, because it is what tells MSstats which
+peptides could have carried a label.
+
+Peptides carrying the label are marked heavy
+(\code{IsotopeLabelType = "H"}), those that could carry it but do not are
+light (\code{"L"}), and those with no labelable residue are \code{NA}.  A
+peptide with two or more labelable residues (counted across all labels
+supplied) may be only partially labeled, which the two-state turnover
+model cannot represent, so it is dropped and the number removed is
+reported.
+
+Defaults to \code{NULL}: turnover mode off, every peptide marked light.}
 }
 \value{
 \code{data.table}
diff --git a/man/dot-formatLogMessage.Rd b/man/dot-formatLogMessage.Rd
index 907fb77b..f3639dd1 100644
--- a/man/dot-formatLogMessage.Rd
+++ b/man/dot-formatLogMessage.Rd
@@ -2,7 +2,7 @@
 % Please edit documentation in R/utils_logging.R
 \name{.formatLogMessage}
 \alias{.formatLogMessage}
-\title{Format a log line as "LEVEL \link{timestamp} message"}
+\title{Format one log line: level, timestamp, message}
 \usage{
 .formatLogMessage(level, ...)
 }