Marker gene validation

[suu-yi]

Implement simple checks on the marker genes before payload construction:

• Check for gene id to gene symbols mapping errors, warning for genes lost in mapping
• Raise value error if marker genes are empty, prevent sending empty marker gene lists
• Validation check if marker genes from adata.uns are gene id like

[qodo-code-review]

Review Summary by Qodo

Add marker gene validation and ID detection checks

✨ Enhancement 🧪 Tests

Walkthroughs

Description

• Add validation for marker gene extraction with lost genes detection
• Raise error when all marker gene lists are empty after mapping
• Detect and warn when rank_genes_groups contain gene IDs instead of symbols
• Implement helper functions to identify gene ID-like patterns

Diagram

flowchart LR
  A["Marker Gene Extraction"] --> B["Check Lost Genes"]
  B --> C["Log Warning if Lost"]
  A --> D["Validate Non-Empty Results"]
  D --> E["Raise Error if Empty"]
  F["Rank Genes Validation"] --> G["Sample Gene Names"]
  G --> H["Detect ID-like Patterns"]
  H --> I["Warn if IDs Found"]

Loading

File Changes

1. cytetype/preprocessing/extraction.py ✨ Enhancement +13/-0

Add marker gene loss tracking and empty validation

• Track and log genes lost during ID-to-name mapping with warning messages
• Add validation to raise ValueError when all marker gene lists are empty
• Provide informative error message about var_names inconsistencies

cytetype/preprocessing/extraction.py

---

2. cytetype/preprocessing/validation.py ✨ Enhancement +22/-0

Add gene ID pattern detection and validation warning

• Sample genes from rank_genes_groups results for pattern analysis
• Implement gene ID-like detection using helper functions _id_like_percentage and
 _is_gene_id_like
• Warn users when gene IDs are detected instead of gene symbols
• Suggest re-running rank_genes_groups with current adata

cytetype/preprocessing/validation.py

---

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0) 🎨 UX Issues (0)

1. Wrong empty-markers error ☑ 🐞 ≡

Description

In extract_marker_genes, the new "all marker gene lists are empty" exception is raised before the
existing "no marker genes found" check, so cases like n_top_genes=0 or empty rank_genes_groups
columns will throw a mapping-mismatch error even though no top genes were extracted.

Code

cytetype/preprocessing/extraction.py[R88-96]

+    if not any(markers.values()):
+        raise ValueError(
+            "All marker gene lists are empty. Gene names in rank_genes_groups "
+            "could not be matched to adata.var_names. This typically happens "
+            "when var_names were changed after rank_genes_groups was run."
+        )
   if not any_genes_found:
       raise ValueError(
           "No marker genes found for any group. This could indicate issues with the "

Evidence

any_genes_found is only set when top_genes is non-empty, but the new emptiness check runs first and
will trigger whenever all extracted marker lists are empty (including when top_genes was empty for
every group). CyteType does not validate that n_top_genes is > 0, so this misleading error path is
user-reachable.

cytetype/preprocessing/extraction.py[69-99]
cytetype/main.py[76-170]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`extract_marker_genes()` raises the new “All marker gene lists are empty…” error before it checks whether any top genes were present at all (`any_genes_found`). This can produce a misleading mapping-mismatch error for cases where no top genes were extracted (e.g., `n_top_genes=0`, or rank_genes_groups produced empty columns).
## Issue Context
- `any_genes_found` only tracks whether `top_genes` was non-empty for at least one group, not whether mapping succeeded.
- The new `if not any(markers.values())` check currently executes before `if not any_genes_found`.
## Fix Focus Areas
- cytetype/preprocessing/extraction.py[69-99]
- cytetype/main.py[76-86]
## Suggested change
1) Validate `n_top_genes >= 1` early (either in `CyteType.__init__` or inside `extract_marker_genes`).
2) Reorder error checks at the end of `extract_marker_genes`:
 - First: if `not any_genes_found`: raise the existing “No marker genes found…” error.
 - Then: if `not any(markers.values())`: raise the new “All marker gene lists are empty…” error.
 - (Optional) If desired, add a more specific message when `any_genes_found` is true but mapping produced empty lists for all clusters.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

2. Noisy gene-ID warning ☑ 🐞 ◔

Description

validate_adata warns whenever rank_genes_groups names look ID-like, but in common setups
rank_genes_groups legitimately returns var_names (often Ensembl IDs) while marker extraction maps
IDs→symbols using gene_symbols_col, making the warning misleading and likely to fire on normal runs.

Code

cytetype/preprocessing/validation.py[R341-361]

+    rank_names = adata.uns[rank_genes_key]["names"]
+    try:
+        sample_genes = [
+            str(g)
+            for field in rank_names.dtype.names[:3]
+            for g in rank_names[field][:20]
+        ]
+    except Exception:
+        sample_genes = []
+
+    if sample_genes:
+        id_pct = _id_like_percentage(sample_genes)
+        if id_pct > 50:
+            examples = [g for g in sample_genes if _is_gene_id_like(g)][:3]
+            logger.warning(
+                f"rank_genes_groups results contain gene IDs rather than gene symbols "
+                f"(e.g. {examples}). This typically happens when var_names were Ensembl "
+                f"IDs at the time rank_genes_groups was run but have since been replaced "
+                f"with gene symbols. Marker gene extraction may fail or produce empty "
+                f"results. Consider re-running sc.tl.rank_genes_groups on the current adata."
+            )

Evidence

The new warning logic checks only the rank_genes_groups 'names' sample and does not consider whether
a gene-symbol mapping column is present. The repo’s own integration fixture uses Ensembl-like var
index values and runs scanpy rank_genes_groups (producing ID-like marker names), while CyteType
passes a resolved gene_symbols_column into validate_adata and later maps those IDs to symbols during
extract_marker_genes—so the warning can be expected even when everything is configured correctly.

cytetype/preprocessing/validation.py[330-362]
cytetype/main.py[157-170]
cytetype/preprocessing/extraction.py[49-79]
tests/conftest.py[34-48]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`validate_adata()` emits a warning when rank_genes_groups marker names look like gene IDs, but it doesn’t account for the (valid) case where IDs are expected and will be mapped to gene symbols via `gene_symbols_col`. This can create noisy warnings and can mislead users (current message suggests var_names were replaced).
## Issue Context
- Rank genes results are sampled from `adata.uns[rank_genes_key]['names']`.
- CyteType commonly runs with Ensembl IDs in `var_names` and a separate gene-symbol column; extraction maps IDs → symbols.
## Fix Focus Areas
- cytetype/preprocessing/validation.py[341-361]
- cytetype/main.py[157-170]
## Suggested change
- Either gate the warning (e.g., only warn when `gene_symbols_col is None`, or when there is evidence of a mismatch such as: `rank_names` are ID-like but `adata.var_names` are not ID-like).
- Alternatively (or additionally), revise the warning text to avoid claiming var_names were replaced; instead state that IDs were detected and marker extraction requires consistent identifiers or a correct mapping column.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

ⓘ The new review experience is currently in Beta. Learn more