speconsense 0.7.3__tar.gz → 0.7.4__tar.gz

{speconsense-0.7.3/speconsense.egg-info → speconsense-0.7.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: speconsense
-Version: 0.7.3
+Version: 0.7.4
 Summary: High-quality clustering and consensus generation for Oxford Nanopore amplicon reads
 Author-email: Josh Walker <joshowalker@yahoo.com>
 License: BSD-3-Clause
@@ -171,7 +171,7 @@ speconsense input.fastq -p herbarium --min-size 10
 ```
 
 **Bundled profiles:**
-- `compressed` — Compress variants into minimal IUPAC consensus sequences (aggressive merging with indels, 20% thresholds, full consensus)
+- `compressed` — Compress variants into minimal IUPAC consensus sequences (aggressive merging with indels, 20% thresholds, full consensus, 20% selection size ratio)
 - `herbarium` — High-recall for degraded DNA/type specimens
 - `largedata` — Experimental settings for large input files
 - `nostalgia` — Simulate older bioinformatics pipelines
@@ -678,6 +678,18 @@ speconsense-summarize --select-strategy diversity --select-max-variants 2
    - Output up to select_max_variants per group
 3. Final output contains representatives from all groups, ensuring both biological diversity (between groups) and appropriate sampling within each biological entity (within groups)
 
+**Selection Size Ratio Filtering:**
+```bash
+speconsense-summarize --select-min-size-ratio 0.2
+```
+- Filters out post-merge variants whose size is too small relative to the largest variant in their group
+- Ratio calculated as `variant_size / largest_size` — must be ≥ threshold to keep
+- Example: `--select-min-size-ratio 0.2` means a variant must have ≥20% the reads of the largest variant in its group
+- Default is 0 (disabled) — all post-merge variants pass through to selection
+- Applied after merging but before variant selection
+- Useful for suppressing noise variants that survived merging but are too small to be meaningful
+- Set to 0.2 in the `compressed` profile to match the 20% calling threshold theme
+
 This two-stage process ensures that distinct biological sequences are preserved as separate groups, while providing control over variant complexity within each group.
 
 ### Customizing FASTA Header Fields
@@ -1059,9 +1071,10 @@ The complete speconsense-summarize workflow operates in this order:
 2. **HAC variant grouping** by sequence identity to separate dissimilar sequences (`--group-identity`); uses single-linkage when overlap merging is enabled
 3. **Group filtering** to limit output groups (`--select-max-groups`)
 4. **Homopolymer-aware MSA-based variant merging** within each group, including **overlap merging** for different-length sequences (`--disable-merging`, `--merge-effort`, `--merge-position-count`, `--merge-indel-length`, `--min-merge-overlap`, `--merge-snp`, `--merge-min-size-ratio`, `--disable-homopolymer-equivalence`)
-5. **Variant selection** within each group (`--select-max-variants`, `--select-strategy`)
-6. **Full consensus generation** (optional) — IUPAC consensus from all pre-merge variants per group (`--enable-full-consensus`)
-7. **Output generation** with customizable header fields (`--fasta-fields`) and full traceability
+5. **Selection size ratio filtering** to remove tiny post-merge variants (`--select-min-size-ratio`)
+6. **Variant selection** within each group (`--select-max-variants`, `--select-strategy`)
+7. **Full consensus generation** (optional) — IUPAC consensus from all pre-merge variants per group (`--enable-full-consensus`)
+8. **Output generation** with customizable header fields (`--fasta-fields`) and full traceability
 
 **Key architectural features**:
 - HAC grouping occurs BEFORE merging to prevent inappropriate merging of dissimilar sequences (e.g., contaminants with primary targets)
@@ -1273,6 +1286,7 @@ usage: speconsense-summarize [-h] [--source SOURCE]
                              [--select-max-groups SELECT_MAX_GROUPS]
                              [--select-max-variants SELECT_MAX_VARIANTS]
                              [--select-strategy {size,diversity}]
+                             [--select-min-size-ratio SELECT_MIN_SIZE_RATIO]
                              [--enable-full-consensus]
                              [--disable-full-consensus]
                              [--scale-threshold SCALE_THRESHOLD] [--threads N]
@@ -1357,6 +1371,10 @@ Selection:
   --select-strategy {size,diversity}, --variant-selection {size,diversity}
                         Variant selection strategy: size or diversity
                         (default: size)
+  --select-min-size-ratio SELECT_MIN_SIZE_RATIO
+                        Minimum size ratio (variant/largest) to include in
+                        output (default: 0 = disabled, e.g. 0.2 for 20%
+                        cutoff)
   --enable-full-consensus
                         Generate a full consensus per variant group
                         representing all variation from pre-merge variants

{speconsense-0.7.3 → speconsense-0.7.4}/README.md

@@ -136,7 +136,7 @@ speconsense input.fastq -p herbarium --min-size 10
 ```
 
 **Bundled profiles:**
-- `compressed` — Compress variants into minimal IUPAC consensus sequences (aggressive merging with indels, 20% thresholds, full consensus)
+- `compressed` — Compress variants into minimal IUPAC consensus sequences (aggressive merging with indels, 20% thresholds, full consensus, 20% selection size ratio)
 - `herbarium` — High-recall for degraded DNA/type specimens
 - `largedata` — Experimental settings for large input files
 - `nostalgia` — Simulate older bioinformatics pipelines
@@ -643,6 +643,18 @@ speconsense-summarize --select-strategy diversity --select-max-variants 2
    - Output up to select_max_variants per group
 3. Final output contains representatives from all groups, ensuring both biological diversity (between groups) and appropriate sampling within each biological entity (within groups)
 
+**Selection Size Ratio Filtering:**
+```bash
+speconsense-summarize --select-min-size-ratio 0.2
+```
+- Filters out post-merge variants whose size is too small relative to the largest variant in their group
+- Ratio calculated as `variant_size / largest_size` — must be ≥ threshold to keep
+- Example: `--select-min-size-ratio 0.2` means a variant must have ≥20% the reads of the largest variant in its group
+- Default is 0 (disabled) — all post-merge variants pass through to selection
+- Applied after merging but before variant selection
+- Useful for suppressing noise variants that survived merging but are too small to be meaningful
+- Set to 0.2 in the `compressed` profile to match the 20% calling threshold theme
+
 This two-stage process ensures that distinct biological sequences are preserved as separate groups, while providing control over variant complexity within each group.
 
 ### Customizing FASTA Header Fields
@@ -1024,9 +1036,10 @@ The complete speconsense-summarize workflow operates in this order:
 2. **HAC variant grouping** by sequence identity to separate dissimilar sequences (`--group-identity`); uses single-linkage when overlap merging is enabled
 3. **Group filtering** to limit output groups (`--select-max-groups`)
 4. **Homopolymer-aware MSA-based variant merging** within each group, including **overlap merging** for different-length sequences (`--disable-merging`, `--merge-effort`, `--merge-position-count`, `--merge-indel-length`, `--min-merge-overlap`, `--merge-snp`, `--merge-min-size-ratio`, `--disable-homopolymer-equivalence`)
-5. **Variant selection** within each group (`--select-max-variants`, `--select-strategy`)
-6. **Full consensus generation** (optional) — IUPAC consensus from all pre-merge variants per group (`--enable-full-consensus`)
-7. **Output generation** with customizable header fields (`--fasta-fields`) and full traceability
+5. **Selection size ratio filtering** to remove tiny post-merge variants (`--select-min-size-ratio`)
+6. **Variant selection** within each group (`--select-max-variants`, `--select-strategy`)
+7. **Full consensus generation** (optional) — IUPAC consensus from all pre-merge variants per group (`--enable-full-consensus`)
+8. **Output generation** with customizable header fields (`--fasta-fields`) and full traceability
 
 **Key architectural features**:
 - HAC grouping occurs BEFORE merging to prevent inappropriate merging of dissimilar sequences (e.g., contaminants with primary targets)
@@ -1238,6 +1251,7 @@ usage: speconsense-summarize [-h] [--source SOURCE]
                              [--select-max-groups SELECT_MAX_GROUPS]
                              [--select-max-variants SELECT_MAX_VARIANTS]
                              [--select-strategy {size,diversity}]
+                             [--select-min-size-ratio SELECT_MIN_SIZE_RATIO]
                              [--enable-full-consensus]
                              [--disable-full-consensus]
                              [--scale-threshold SCALE_THRESHOLD] [--threads N]
@@ -1322,6 +1336,10 @@ Selection:
   --select-strategy {size,diversity}, --variant-selection {size,diversity}
                         Variant selection strategy: size or diversity
                         (default: size)
+  --select-min-size-ratio SELECT_MIN_SIZE_RATIO
+                        Minimum size ratio (variant/largest) to include in
+                        output (default: 0 = disabled, e.g. 0.2 for 20%
+                        cutoff)
   --enable-full-consensus
                         Generate a full consensus per variant group
                         representing all variation from pre-merge variants

{speconsense-0.7.3 → speconsense-0.7.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "speconsense"
-version = "0.7.3"
+version = "0.7.4"
 description = "High-quality clustering and consensus generation for Oxford Nanopore amplicon reads"
 readme = "README.md"
 requires-python = ">=3.8"

{speconsense-0.7.3 → speconsense-0.7.4}/speconsense/__init__.py

@@ -5,7 +5,7 @@ A Python tool for experimental clustering and consensus generation as an alterna
 in the fungal DNA barcoding pipeline.
 """
 
-__version__ = "0.7.3"
+__version__ = "0.7.4"
 __author__ = "Josh Walker"
 __email__ = "joshowalker@yahoo.com"
 

{speconsense-0.7.3 → speconsense-0.7.4}/speconsense/profiles/__init__.py

@@ -103,6 +103,7 @@ VALID_SUMMARIZE_KEYS = {
     "select-max-groups",
     "select-max-variants",
     "select-strategy",
+    "select-min-size-ratio",
     "enable-full-consensus",
     # Processing
     "scale-threshold",

{speconsense-0.7.3 → speconsense-0.7.4}/speconsense/profiles/compressed.yaml

@@ -23,5 +23,6 @@ speconsense-summarize:
   merge-indel-length: 5         # Merge indels up to 5bp
   merge-position-count: 10      # Allow up to 10 variant positions in a merge
   merge-min-size-ratio: 0.2     # Match 20% calling threshold
+  select-min-size-ratio: 0.2    # Match 20% calling threshold
   min-merge-overlap: 0          # Disable partial overlap merging
   enable-full-consensus: true   # Include full IUPAC consensus per group

{speconsense-0.7.3 → speconsense-0.7.4}/speconsense/profiles/example.yaml

@@ -91,6 +91,7 @@ speconsense-summarize:
   # select-max-groups: -1         # Max groups to output (-1 = no limit)
   # select-max-variants: -1       # Max variants per group (-1 = no limit)
   # select-strategy: size         # Selection strategy: size or diversity
+  # select-min-size-ratio: 0    # Min size ratio to include variant (0 = disabled)
 
   # --- Processing ---
   # threads: 0                    # Max threads (0 = auto-detect)

{speconsense-0.7.3 → speconsense-0.7.4}/speconsense/summarize/cli.py

@@ -169,6 +169,9 @@ def parse_arguments():
     selection_group.add_argument("--select-strategy", "--variant-selection",
                                  dest="select_strategy", choices=["size", "diversity"], default="size",
                                  help="Variant selection strategy: size or diversity (default: size)")
+    selection_group.add_argument("--select-min-size-ratio", type=float, default=0,
+                                 help="Minimum size ratio (variant/largest) to include in output "
+                                      "(default: 0 = disabled, e.g. 0.2 for 20%% cutoff)")
     selection_group.add_argument("--enable-full-consensus", action="store_true",
                                  help="Generate a full consensus per variant group representing all variation "
                                       "from pre-merge variants (gaps never win)")
@@ -359,6 +362,18 @@ def process_single_specimen(file_consensuses: List[ConsensusInfo],
     for group_idx, (group_id, group_members) in enumerate(sorted_groups):
         final_group_name = group_idx + 1
 
+        # Apply select-min-size-ratio filter
+        if args.select_min_size_ratio > 0 and len(group_members) > 1:
+            largest_size = max(v.size for v in group_members)
+            filtered = [v for v in group_members
+                        if (v.size / largest_size) >= args.select_min_size_ratio]
+            if len(filtered) < len(group_members):
+                filtered_count = len(group_members) - len(filtered)
+                logging.debug(f"Group {group_idx + 1}: filtered out {filtered_count} "
+                              f"variants with size ratio < {args.select_min_size_ratio} "
+                              f"relative to largest (size={largest_size})")
+                group_members = filtered
+
         # Select variants for this group
         selected_variants = select_variants(group_members, args.select_max_variants, args.select_strategy, group_number=final_group_name)
 
@@ -380,6 +395,17 @@ def process_single_specimen(file_consensuses: List[ConsensusInfo],
         # Generate full consensus from PRE-MERGE variants
         if getattr(args, 'enable_full_consensus', False):
             pre_merge_variants = variant_groups[group_id]
+
+            # Apply size-ratio filter (same as merge pipeline)
+            if args.merge_min_size_ratio > 0 and len(pre_merge_variants) > 1:
+                largest_size = max(v.size for v in pre_merge_variants)
+                filtered = [v for v in pre_merge_variants
+                            if (v.size / largest_size) >= args.merge_min_size_ratio]
+                if len(filtered) < len(pre_merge_variants):
+                    filtered_count = len(pre_merge_variants) - len(filtered)
+                    logging.debug(f"Full consensus: filtered out {filtered_count} variants with size ratio < {args.merge_min_size_ratio} relative to largest (size={largest_size})")
+                    pre_merge_variants = filtered
+
             specimen_base = selected_variants[0].sample_name.rsplit('-c', 1)[0]
             full_name = f"{specimen_base}-{group_idx + 1}.full"
 
@@ -450,6 +476,7 @@ def main():
     logging.info(f"  --select-max-variants: {args.select_max_variants}")
     logging.info(f"  --select-max-groups: {args.select_max_groups}")
     logging.info(f"  --select-strategy: {args.select_strategy}")
+    logging.info(f"  --select-min-size-ratio: {args.select_min_size_ratio}")
     logging.info(f"  --enable-full-consensus: {args.enable_full_consensus}")
     logging.info(f"  --log-level: {args.log_level}")
     logging.info("")

{speconsense-0.7.3 → speconsense-0.7.4/speconsense.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: speconsense
-Version: 0.7.3
+Version: 0.7.4
 Summary: High-quality clustering and consensus generation for Oxford Nanopore amplicon reads
 Author-email: Josh Walker <joshowalker@yahoo.com>
 License: BSD-3-Clause
@@ -171,7 +171,7 @@ speconsense input.fastq -p herbarium --min-size 10
 ```
 
 **Bundled profiles:**
-- `compressed` — Compress variants into minimal IUPAC consensus sequences (aggressive merging with indels, 20% thresholds, full consensus)
+- `compressed` — Compress variants into minimal IUPAC consensus sequences (aggressive merging with indels, 20% thresholds, full consensus, 20% selection size ratio)
 - `herbarium` — High-recall for degraded DNA/type specimens
 - `largedata` — Experimental settings for large input files
 - `nostalgia` — Simulate older bioinformatics pipelines
@@ -678,6 +678,18 @@ speconsense-summarize --select-strategy diversity --select-max-variants 2
    - Output up to select_max_variants per group
 3. Final output contains representatives from all groups, ensuring both biological diversity (between groups) and appropriate sampling within each biological entity (within groups)
 
+**Selection Size Ratio Filtering:**
+```bash
+speconsense-summarize --select-min-size-ratio 0.2
+```
+- Filters out post-merge variants whose size is too small relative to the largest variant in their group
+- Ratio calculated as `variant_size / largest_size` — must be ≥ threshold to keep
+- Example: `--select-min-size-ratio 0.2` means a variant must have ≥20% the reads of the largest variant in its group
+- Default is 0 (disabled) — all post-merge variants pass through to selection
+- Applied after merging but before variant selection
+- Useful for suppressing noise variants that survived merging but are too small to be meaningful
+- Set to 0.2 in the `compressed` profile to match the 20% calling threshold theme
+
 This two-stage process ensures that distinct biological sequences are preserved as separate groups, while providing control over variant complexity within each group.
 
 ### Customizing FASTA Header Fields
@@ -1059,9 +1071,10 @@ The complete speconsense-summarize workflow operates in this order:
 2. **HAC variant grouping** by sequence identity to separate dissimilar sequences (`--group-identity`); uses single-linkage when overlap merging is enabled
 3. **Group filtering** to limit output groups (`--select-max-groups`)
 4. **Homopolymer-aware MSA-based variant merging** within each group, including **overlap merging** for different-length sequences (`--disable-merging`, `--merge-effort`, `--merge-position-count`, `--merge-indel-length`, `--min-merge-overlap`, `--merge-snp`, `--merge-min-size-ratio`, `--disable-homopolymer-equivalence`)
-5. **Variant selection** within each group (`--select-max-variants`, `--select-strategy`)
-6. **Full consensus generation** (optional) — IUPAC consensus from all pre-merge variants per group (`--enable-full-consensus`)
-7. **Output generation** with customizable header fields (`--fasta-fields`) and full traceability
+5. **Selection size ratio filtering** to remove tiny post-merge variants (`--select-min-size-ratio`)
+6. **Variant selection** within each group (`--select-max-variants`, `--select-strategy`)
+7. **Full consensus generation** (optional) — IUPAC consensus from all pre-merge variants per group (`--enable-full-consensus`)
+8. **Output generation** with customizable header fields (`--fasta-fields`) and full traceability
 
 **Key architectural features**:
 - HAC grouping occurs BEFORE merging to prevent inappropriate merging of dissimilar sequences (e.g., contaminants with primary targets)
@@ -1273,6 +1286,7 @@ usage: speconsense-summarize [-h] [--source SOURCE]
                              [--select-max-groups SELECT_MAX_GROUPS]
                              [--select-max-variants SELECT_MAX_VARIANTS]
                              [--select-strategy {size,diversity}]
+                             [--select-min-size-ratio SELECT_MIN_SIZE_RATIO]
                              [--enable-full-consensus]
                              [--disable-full-consensus]
                              [--scale-threshold SCALE_THRESHOLD] [--threads N]
@@ -1357,6 +1371,10 @@ Selection:
   --select-strategy {size,diversity}, --variant-selection {size,diversity}
                         Variant selection strategy: size or diversity
                         (default: size)
+  --select-min-size-ratio SELECT_MIN_SIZE_RATIO
+                        Minimum size ratio (variant/largest) to include in
+                        output (default: 0 = disabled, e.g. 0.2 for 20%
+                        cutoff)
   --enable-full-consensus
                         Generate a full consensus per variant group
                         representing all variation from pre-merge variants

{speconsense-0.7.3 → speconsense-0.7.4}/tests/test_summarize.py

@@ -520,6 +520,114 @@ class TestFullConsensus:
         assert result.rid == 0.95
 
 
+    def test_full_consensus_filters_small_variants(self):
+        """Integration test: merge_min_size_ratio filters small variants from full consensus."""
+        temp_dir = tempfile.mkdtemp()
+        source_dir = os.path.join(temp_dir, "clusters")
+        summary_dir = os.path.join(temp_dir, "__Summary__")
+        os.makedirs(source_dir)
+
+        try:
+            # Two similar sequences (1 SNP at position 12: G vs A)
+            # Very different sizes so the small one is filtered by merge_min_size_ratio
+            seq_large = "ATCGATCGATCGATCGATCGATCG"  # G at position 12
+            seq_small = "ATCGATCGATCAATCGATCGATCG"  # A at position 12
+
+            fasta_content = f""">test-c1 size=100 ric=100 primers=test
+{seq_large}
+>test-c2 size=5 ric=5 primers=test
+{seq_small}
+"""
+            fasta_file = os.path.join(source_dir, "test-all.fasta")
+            with open(fasta_file, 'w') as f:
+                f.write(fasta_content)
+
+            # merge-min-size-ratio 0.1 filters 5/100=0.05 from full consensus
+            result = subprocess.run(
+                [
+                    "speconsense-summarize",
+                    "--source", source_dir,
+                    "--summary-dir", summary_dir,
+                    "--min-ric", "3",
+                    "--enable-full-consensus",
+                    "--merge-min-size-ratio", "0.1",
+                    "--disable-merging",
+                    "--min-merge-overlap", "0",
+                ],
+                capture_output=True,
+                text=True
+            )
+
+            assert result.returncode == 0, f"speconsense-summarize failed: {result.stderr}"
+
+            output_fasta = os.path.join(summary_dir, "summary.fasta")
+            output_sequences = list(SeqIO.parse(output_fasta, "fasta"))
+
+            full_seqs = [s for s in output_sequences if '.full' in s.id]
+            assert len(full_seqs) == 1, f"Expected 1 .full sequence, got {len(full_seqs)}"
+
+            # Small variant was filtered — .full should be the large variant only (no IUPAC)
+            full_seq_str = str(full_seqs[0].seq)
+            assert full_seq_str == seq_large, \
+                f"Expected large variant sequence, got {full_seq_str}"
+
+        finally:
+            shutil.rmtree(temp_dir)
+
+    def test_full_consensus_no_filter_when_disabled(self):
+        """Integration test: merge_min_size_ratio=0 preserves all variants in full consensus."""
+        temp_dir = tempfile.mkdtemp()
+        source_dir = os.path.join(temp_dir, "clusters")
+        summary_dir = os.path.join(temp_dir, "__Summary__")
+        os.makedirs(source_dir)
+
+        try:
+            # Same sequences as above — 1 SNP at position 12 (G vs A)
+            seq_large = "ATCGATCGATCGATCGATCGATCG"
+            seq_small = "ATCGATCGATCAATCGATCGATCG"
+
+            fasta_content = f""">test-c1 size=100 ric=100 primers=test
+{seq_large}
+>test-c2 size=5 ric=5 primers=test
+{seq_small}
+"""
+            fasta_file = os.path.join(source_dir, "test-all.fasta")
+            with open(fasta_file, 'w') as f:
+                f.write(fasta_content)
+
+            # merge-min-size-ratio 0 disables filtering — both contribute to .full
+            result = subprocess.run(
+                [
+                    "speconsense-summarize",
+                    "--source", source_dir,
+                    "--summary-dir", summary_dir,
+                    "--min-ric", "3",
+                    "--enable-full-consensus",
+                    "--merge-min-size-ratio", "0",
+                    "--disable-merging",
+                    "--min-merge-overlap", "0",
+                ],
+                capture_output=True,
+                text=True
+            )
+
+            assert result.returncode == 0, f"speconsense-summarize failed: {result.stderr}"
+
+            output_fasta = os.path.join(summary_dir, "summary.fasta")
+            output_sequences = list(SeqIO.parse(output_fasta, "fasta"))
+
+            full_seqs = [s for s in output_sequences if '.full' in s.id]
+            assert len(full_seqs) == 1, f"Expected 1 .full sequence, got {len(full_seqs)}"
+
+            # Both variants contribute — SNP position should be IUPAC R (A/G)
+            full_seq_str = str(full_seqs[0].seq)
+            assert "R" in full_seq_str, \
+                f"Expected IUPAC R (A/G) in full consensus, got {full_seq_str}"
+
+        finally:
+            shutil.rmtree(temp_dir)
+
+
 class TestFieldRegexFullConsensus:
     """Tests for GroupField and VariantField regex handling of .full names."""
 
@@ -578,6 +686,101 @@ class TestFieldRegexFullConsensus:
         assert field.format_value(cons) == "variant=v1"
 
 
+class TestSelectMinSizeRatio:
+    """Tests for --select-min-size-ratio filtering."""
+
+    def test_select_min_size_ratio_filters_small_variants(self):
+        """Integration test: --select-min-size-ratio 0.1 filters out tiny variants."""
+        temp_dir = tempfile.mkdtemp()
+        source_dir = os.path.join(temp_dir, "clusters")
+        summary_dir = os.path.join(temp_dir, "__Summary__")
+        os.makedirs(source_dir)
+
+        try:
+            seq1 = "ATCGATCGATCGATCGATCGATCG"
+            seq2 = "ATCGATCGATCAATCGATCGATCG"  # One SNP — different enough to not merge
+
+            fasta_content = f""">test-c1 size=100 ric=100 primers=test
+{seq1}
+>test-c2 size=3 ric=3 primers=test
+{seq2}
+"""
+            fasta_file = os.path.join(source_dir, "test-all.fasta")
+            with open(fasta_file, 'w') as f:
+                f.write(fasta_content)
+
+            result = subprocess.run(
+                [
+                    "speconsense-summarize",
+                    "--source", source_dir,
+                    "--summary-dir", summary_dir,
+                    "--min-ric", "3",
+                    "--select-min-size-ratio", "0.1",
+                    "--disable-merging",
+                ],
+                capture_output=True,
+                text=True
+            )
+
+            assert result.returncode == 0, f"speconsense-summarize failed: {result.stderr}"
+
+            output_fasta = os.path.join(summary_dir, "summary.fasta")
+            output_sequences = list(SeqIO.parse(output_fasta, "fasta"))
+
+            # Only the large variant should remain (3/100 = 0.03 < 0.1)
+            assert len(output_sequences) == 1, \
+                f"Expected 1 sequence after filtering, got {len(output_sequences)}"
+            assert "size=100" in output_sequences[0].description
+
+        finally:
+            shutil.rmtree(temp_dir)
+
+    def test_select_min_size_ratio_disabled_preserves_all(self):
+        """Integration test: --select-min-size-ratio 0 preserves all variants."""
+        temp_dir = tempfile.mkdtemp()
+        source_dir = os.path.join(temp_dir, "clusters")
+        summary_dir = os.path.join(temp_dir, "__Summary__")
+        os.makedirs(source_dir)
+
+        try:
+            seq1 = "ATCGATCGATCGATCGATCGATCG"
+            seq2 = "ATCGATCGATCAATCGATCGATCG"  # One SNP
+
+            fasta_content = f""">test-c1 size=100 ric=100 primers=test
+{seq1}
+>test-c2 size=3 ric=3 primers=test
+{seq2}
+"""
+            fasta_file = os.path.join(source_dir, "test-all.fasta")
+            with open(fasta_file, 'w') as f:
+                f.write(fasta_content)
+
+            result = subprocess.run(
+                [
+                    "speconsense-summarize",
+                    "--source", source_dir,
+                    "--summary-dir", summary_dir,
+                    "--min-ric", "3",
+                    "--select-min-size-ratio", "0",
+                    "--disable-merging",
+                ],
+                capture_output=True,
+                text=True
+            )
+
+            assert result.returncode == 0, f"speconsense-summarize failed: {result.stderr}"
+
+            output_fasta = os.path.join(summary_dir, "summary.fasta")
+            output_sequences = list(SeqIO.parse(output_fasta, "fasta"))
+
+            # Both variants should be preserved
+            assert len(output_sequences) == 2, \
+                f"Expected 2 sequences with ratio=0, got {len(output_sequences)}"
+
+        finally:
+            shutil.rmtree(temp_dir)
+
+
 class TestFullConsensusIntegration:
     """Integration test for --enable-full-consensus."""