Added DFS with pruning

CHANGELOG.md

@@ -5,6 +5,25 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.2.0] - 26-03-2026
+
+### Added
+
+- **Enriched CSV output for post hoc analysis** (`designer/multiplexpanel.py`, `designer/primer.py`): `selected_multiplex.csv`, `top_panels.csv`, and `candidate_pairs.csv` now include 12 additional per-primer and per-amplicon columns: `Forward/Reverse_Self_Any_Th`, `Forward/Reverse_Self_End_Th`, `Forward/Reverse_Hairpin_Th`, `Forward/Reverse_End_Stability`, `Forward/Reverse_Penalty`, `Amplicon_GC`, and `Num_Candidate_Pairs`. These expose design-time thermodynamic features needed to correlate primer properties with wet-lab validation coverage.
+- **Per-assay cross-dimer contribution** (`designer/multiplexpanel.py`): `selected_multiplex.csv` and `top_panels.csv` now include a `Cross_Dimer_Contribution` column quantifying the sum of cross-dimer interaction scores between each assay's primers and all other primers in the multiplex. Mirrors the scoring logic in `selector/cost.py` with cached pairwise alignments.
+
+- **Primer-level cross-dimer matrix** (`reporting/qc.py`): `generate_panel_qc()` now computes a full primer-to-primer dimer matrix for the selected multiplex, covering all C(2n, 2) individual primer interactions (including intra-junction pairs). The matrix is returned alongside the existing junction-level summary. Tail sequences (`forward_tail`, `reverse_tail`) are now included in the dimer alignment for realistic scoring. A new `save_primer_dimer_matrix_csv()` function exports the symmetric matrix as a CSV file (`primer_dimer_matrix.csv`), written automatically by the pipeline.
+- **Primer-level dimer heatmap in HTML report** (`reporting/templates/panel_report.html.j2`): New interactive Plotly heatmap section showing individual primer-to-primer dimer scores, displayed below the existing junction-level heatmap.
+
+### Fixed
+
+- **Cross-reactivity heatmap not rendering** (`reporting/templates/plotly.min.js.gz`): The bundled Plotly.js was the "basic" partial build which only includes scatter, bar, and pie trace types. The heatmap trace type was missing, causing the cross-reactivity heatmap to silently fail. Replaced with the full Plotly.js v2.35.2 bundle.
+
+- **Target dropout for DFS selector** (`selector/selectors.py`): The DFS selector can now optionally drop targets whose primers cause extreme cross-dimer interactions that "poison" the panel. Enabled via `allow_target_dropping: true` in the multiplex picker config. An adaptive dropout penalty is computed from the greedy seed's marginal cross-dimer costs at a configurable percentile (`dropout_stringency`, default 0.8), so only true outliers are removed. A hard floor prevents excessive dropping: `max(minimum_plexity, ceil(min_target_fraction * n_input))`, clamped to the actual input count.
+- **New config fields** (`config.py`, preset JSON files): `allow_target_dropping` (bool, default false), `dropout_stringency` (float 0-1, default 0.8), `min_target_fraction` (float 0-1, default 0.8).
+- **Dropped targets in output** (`designer/multiplexpanel.py`): `panel_summary.json` includes a `dropped_targets` list. `top_panels.csv` includes `Num_Dropped` and `Dropped_Targets` columns. Dropped targets are also logged as warnings and surfaced in CLI output.
+- **Plexity clamping** (`pipeline.py`): When the number of input targets is less than the configured `maximum_plexity` or `minimum_plexity`, the effective plexity values are clamped to the actual input count, preventing nonsensical constraints.
+
 ## [1.1.0] - 13-03-2026
 
 ### Changed

docker/DOCKERFILE

@@ -1,6 +1,6 @@
 ARG BLAST_VERSION=2.17.0
 ARG BCFTOOLS_VERSION=1.23
-ARG PLEXUS_VERSION=1.0.0
+ARG PLEXUS_VERSION=1.2.0
 
 # ── Stage 1: Python venv builder ─────────────────────────────────────────────
 FROM ghcr.io/astral-sh/uv:python3.13-bookworm-slim AS builder
@@ -43,18 +43,18 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 
 # Download and extract exact BLAST+ binaries (NCBI prebuilt)
 RUN wget -qO /tmp/blast.tar.gz \
-      "https://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/${BLAST_VERSION}/ncbi-blast-${BLAST_VERSION}+-x64-linux.tar.gz" \
+    "https://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/${BLAST_VERSION}/ncbi-blast-${BLAST_VERSION}+-x64-linux.tar.gz" \
     && tar -xzf /tmp/blast.tar.gz -C /tmp \
     && mkdir -p /tools/bin \
     && cp /tmp/ncbi-blast-${BLAST_VERSION}+/bin/blastn \
-          /tmp/ncbi-blast-${BLAST_VERSION}+/bin/makeblastdb \
-          /tmp/ncbi-blast-${BLAST_VERSION}+/bin/blast_formatter \
-          /tools/bin/ \
+    /tmp/ncbi-blast-${BLAST_VERSION}+/bin/makeblastdb \
+    /tmp/ncbi-blast-${BLAST_VERSION}+/bin/blast_formatter \
+    /tools/bin/ \
     && rm -rf /tmp/blast.tar.gz /tmp/ncbi-blast-${BLAST_VERSION}+
 
 # Compile exact bcftools from source
 RUN wget -qO /tmp/bcftools.tar.bz2 \
-      "https://github.com/samtools/bcftools/releases/download/${BCFTOOLS_VERSION}/bcftools-${BCFTOOLS_VERSION}.tar.bz2" \
+    "https://github.com/samtools/bcftools/releases/download/${BCFTOOLS_VERSION}/bcftools-${BCFTOOLS_VERSION}.tar.bz2" \
     && tar -xjf /tmp/bcftools.tar.bz2 -C /tmp \
     && cd /tmp/bcftools-${BCFTOOLS_VERSION} \
     && make -j$(nproc) \
@@ -70,12 +70,12 @@ ARG PLEXUS_VERSION
 
 # Bake version and audit metadata into the image
 LABEL org.opencontainers.image.title="plexus" \
-      org.opencontainers.image.version="${PLEXUS_VERSION}" \
-      org.opencontainers.image.description="Multiplex PCR primer panel designer" \
-      org.opencontainers.image.licenses="GPL-2.0-or-later" \
-      dev.plexus.blast_version="${BLAST_VERSION}" \
-      dev.plexus.bcftools_version="${BCFTOOLS_VERSION}" \
-      dev.plexus.compliance_mode="true"
+    org.opencontainers.image.version="${PLEXUS_VERSION}" \
+    org.opencontainers.image.description="Multiplex PCR primer panel designer" \
+    org.opencontainers.image.licenses="GPL-2.0-or-later" \
+    dev.plexus.blast_version="${BLAST_VERSION}" \
+    dev.plexus.bcftools_version="${BCFTOOLS_VERSION}" \
+    dev.plexus.compliance_mode="true"
 
 # Non-root runtime user
 RUN useradd --system --create-home --shell /bin/bash plexus

src/plexus/config.py

@@ -183,6 +183,31 @@ class MultiplexPickerParameters(BaseModel):
     plexity_wt_gt: float = Field(default=1.0, ge=0.0)
 
     force_plexity: bool = Field(default=False)
+
+    allow_target_dropping: bool = Field(
+        default=False,
+        description="Allow the DFS selector to drop targets with extreme cross-dimer toxicity.",
+    )
+    dropout_stringency: float = Field(
+        default=0.8,
+        ge=0.0,
+        le=1.0,
+        description=(
+            "Percentile (0-1) of marginal cross-dimer distribution used to set "
+            "the dropout penalty. Higher values = stricter = fewer drops. "
+            "0.8 means only targets worse than the 80th percentile are drop candidates."
+        ),
+    )
+    min_target_fraction: float = Field(
+        default=0.8,
+        ge=0.0,
+        le=1.0,
+        description=(
+            "Minimum fraction of input targets to retain. "
+            "Effective floor = max(minimum_plexity, ceil(min_target_fraction * n_input))."
+        ),
+    )
+
     allow_split_panel: bool = Field(default=False)
     max_splits: int = Field(default=2, ge=1, le=10)
 

src/plexus/data/designer_default_config.json

@@ -94,6 +94,9 @@
         "plexity_wt_lt": 1.0,
         "plexity_wt_gt": 1.0,
         "force_plexity": false,
+        "allow_target_dropping": false,
+        "dropout_stringency": 0.8,
+        "min_target_fraction": 0.8,
         "allow_split_panel": false,
         "max_splits": 2,
         "wt_pair_penalty": 1.0,

src/plexus/data/designer_lenient_config.json

@@ -94,6 +94,9 @@
         "plexity_wt_lt": 1.0,
         "plexity_wt_gt": 1.0,
         "force_plexity": false,
+        "allow_target_dropping": false,
+        "dropout_stringency": 0.8,
+        "min_target_fraction": 0.8,
         "allow_split_panel": false,
         "max_splits": 2,
         "wt_pair_penalty": 1.0,

src/plexus/designer/multiplexpanel.py

@@ -833,6 +833,20 @@ def save_candidate_pairs_to_csv(self, file_path: str):
                     "Off_Target_Count": len(pair.off_target_products),
                     "Specificity_Checked": pair.specificity_checked,
                     "Selected": pair.selected,
+                    "Forward_Self_Any_Th": pair.forward.self_any_th,
+                    "Reverse_Self_Any_Th": pair.reverse.self_any_th,
+                    "Forward_Self_End_Th": pair.forward.self_end_th,
+                    "Reverse_Self_End_Th": pair.reverse.self_end_th,
+                    "Forward_Hairpin_Th": pair.forward.hairpin_th,
+                    "Reverse_Hairpin_Th": pair.reverse.hairpin_th,
+                    "Forward_End_Stability": pair.forward.end_stability,
+                    "Reverse_End_Stability": pair.reverse.end_stability,
+                    "Forward_Penalty": pair.forward.penalty,
+                    "Reverse_Penalty": pair.reverse.penalty,
+                    "Amplicon_GC": pair.amplicon_gc,
+                    "Num_Candidate_Pairs": (
+                        len(junction.primer_pairs) if junction.primer_pairs else 0
+                    ),
                 }
                 data.append(row)
 
@@ -901,6 +915,20 @@ def _build_enriched_pair_row(self, junction: Junction, pair: PrimerPair) -> dict
             "SNP_Penalty": pair.snp_penalty,
             "Forward_SNP_Count": pair.forward.snp_count,
             "Reverse_SNP_Count": pair.reverse.snp_count,
+            "Forward_Self_Any_Th": pair.forward.self_any_th,
+            "Reverse_Self_Any_Th": pair.reverse.self_any_th,
+            "Forward_Self_End_Th": pair.forward.self_end_th,
+            "Reverse_Self_End_Th": pair.reverse.self_end_th,
+            "Forward_Hairpin_Th": pair.forward.hairpin_th,
+            "Reverse_Hairpin_Th": pair.reverse.hairpin_th,
+            "Forward_End_Stability": pair.forward.end_stability,
+            "Reverse_End_Stability": pair.reverse.end_stability,
+            "Forward_Penalty": pair.forward.penalty,
+            "Reverse_Penalty": pair.reverse.penalty,
+            "Amplicon_GC": pair.amplicon_gc,
+            "Num_Candidate_Pairs": (
+                len(junction.primer_pairs) if junction.primer_pairs else 0
+            ),
         }
 
     def _junction_for_pair(self, pair: PrimerPair) -> Junction | None:
@@ -924,13 +952,41 @@ def save_selected_multiplex_csv(self, file_path: str, selected_pairs: list) -> N
             logger.warning("No selected pairs to save.")
             return
 
+        # Pre-compute cross-dimer scores for per-pair attribution
+        dimer_predictor = PrimerDimerPredictor()
+        dimer_cache: dict[tuple[str, str], float] = {}
+        all_primers: dict[str, list[tuple[str, str]]] = {}
+        for pair in selected_pairs:
+            all_primers[pair.pair_id] = [
+                (pair.forward.seq, pair.forward.name),
+                (pair.reverse.seq, pair.reverse.name),
+            ]
+
         data = []
         for pair in selected_pairs:
             junction = self._junction_for_pair(pair)
             if junction is None:
                 logger.warning(f"Could not find junction for pair {pair.pair_id}")
                 continue
-            data.append(self._build_enriched_pair_row(junction, pair))
+            row = self._build_enriched_pair_row(junction, pair)
+
+            contribution = 0.0
+            for other in selected_pairs:
+                if other.pair_id == pair.pair_id:
+                    continue
+                for seq_a, name_a in all_primers[pair.pair_id]:
+                    for seq_b, name_b in all_primers[other.pair_id]:
+                        key = tuple(sorted((seq_a, seq_b)))
+                        if key not in dimer_cache:
+                            dimer_predictor.set_primers(seq_a, seq_b, name_a, name_b)
+                            dimer_predictor.align()
+                            dimer_cache[key] = dimer_predictor.score or 0.0
+                        score = dimer_cache[key]
+                        if score < 0:
+                            contribution += abs(score)
+            row["Cross_Dimer_Contribution"] = round(contribution, 4)
+
+            data.append(row)
 
         df = pd.DataFrame(data)
         df.to_csv(file_path, index=False)
@@ -948,19 +1004,56 @@ def save_top_panels_csv(self, file_path: str, solutions: list) -> None:
             return
 
         pair_lookup = self.build_pair_lookup()
+        dimer_predictor = PrimerDimerPredictor()
+        dimer_cache: dict[tuple[str, str], float] = {}
         data = []
 
         for rank, solution in enumerate(solutions, start=1):
+            # Collect primers for this solution's pairs
+            sol_primers: dict[str, list[tuple[str, str]]] = {}
+            sol_pairs: list[PrimerPair] = []
             for pair_id in solution.primer_pairs:
                 pair = pair_lookup.get(pair_id)
                 if pair is None:
                     continue
+                sol_pairs.append(pair)
+                sol_primers[pair_id] = [
+                    (pair.forward.seq, pair.forward.name),
+                    (pair.reverse.seq, pair.reverse.name),
+                ]
+
+            for pair in sol_pairs:
                 junction = self._junction_for_pair(pair)
                 if junction is None:
                     continue
                 row = self._build_enriched_pair_row(junction, pair)
+
+                contribution = 0.0
+                for other in sol_pairs:
+                    if other.pair_id == pair.pair_id:
+                        continue
+                    for seq_a, name_a in sol_primers[pair.pair_id]:
+                        for seq_b, name_b in sol_primers[other.pair_id]:
+                            key = tuple(sorted((seq_a, seq_b)))
+                            if key not in dimer_cache:
+                                dimer_predictor.set_primers(
+                                    seq_a, seq_b, name_a, name_b
+                                )
+                                dimer_predictor.align()
+                                dimer_cache[key] = dimer_predictor.score or 0.0
+                            score = dimer_cache[key]
+                            if score < 0:
+                                contribution += abs(score)
+                row["Cross_Dimer_Contribution"] = round(contribution, 4)
+
                 row["Solution_Rank"] = rank
                 row["Solution_Cost"] = round(solution.cost, 4)
+                row["Num_Dropped"] = len(solution.dropped_targets)
+                row["Dropped_Targets"] = (
+                    "; ".join(solution.dropped_targets)
+                    if solution.dropped_targets
+                    else ""
+                )
                 data.append(row)
 
         df = pd.DataFrame(data)
@@ -1054,6 +1147,9 @@ def save_panel_summary_json(
             "num_junctions": len(self.junctions),
             "num_candidate_pairs": total_pairs,
             "num_selected_pairs": len(pipeline_result.selected_pairs),
+            "dropped_targets": pipeline_result.multiplex_solutions[0].dropped_targets
+            if pipeline_result.multiplex_solutions
+            else [],
             "best_multiplex_cost": round(pipeline_result.multiplex_solutions[0].cost, 4)
             if pipeline_result.multiplex_solutions
             else None,

src/plexus/designer/primer.py

@@ -71,6 +71,15 @@ class PrimerPair:
     snp_count: int = 0
     snp_penalty: float = 0.0
 
+    @property
+    def amplicon_gc(self) -> float | None:
+        """GC content of the amplicon sequence as a percentage."""
+        seq = self.amplicon_sequence
+        if not seq:
+            return None
+        gc_count = seq.upper().count("G") + seq.upper().count("C")
+        return round(100.0 * gc_count / len(seq), 2)
+
     @staticmethod
     def calculate_primer_pair_penalty_th(
         primer_left_penalty,