← All runs

cBioPortalChat benchmark · 20260925-1407

14 questions (14 with a reference answer, link or rubric) · target https://beta.chat.cbioportal.org (agent agent_OHVSJI9Gd6gwsDnFSL-Xl) · judge us.anthropic.claude-sonnet-4-6 ($0.175)

runner claude-code (headless Claude Code with the agent's prompt and MCP servers — compare with other claude-code runs, not with agents-api runs) · prompt: beta agent agent_OHVSJI9Gd6gwsDnFSL-Xl, 5dbe6fa4ced1, 14476 chars, agent last updated 2026-09-24 13:11 UTC · database MCP claude.ai cBioPortal MCP, navigator https://mcp.cbioportal.org/navigator/mcp

cBioPortal v7.1.2 (DB schema 3.0.0, hgnc_v7_2025.10.7) · cbioportal-mcp (sha256:a06ebc2283e0) · navigator 1.0.0 (sha256:4135b03b8257) · 2.1.282 (Claude Code)

Headline

Precision: pass rate on questions the model attempted. Coverage: share it attempted rather than declined. Costs are what these tokens would cost at Anthropic list prices; this run was answered on a Claude subscription and billed nothing per token.

Outcomes

Pass rate by track

Pass rate by category

Tokens and cost

Latency and tool use

Objective checks

Questions

In MSK-IMPACT (msk_impact_50k_2026), **KRAS G12C** by cancer type (samples profiled for KRAS):

- **Non-Small Cell Lung Cancer**: 922 / 7,867 (11.7%)
- **Appendiceal Cancer**: 12 / 274 (4.4%)
- **Small Bowel Cancer**: 7 / 161 (4.3%)
- **Colorectal Cancer**: 158 / 5,277 (3.0%)
- **Cancer of Unknown Primary**: 48 / 1,581 (3.0%)
- **Ampullary Cancer**: 4 / 134 (3.0%)
- **Vaginal Cancer**: 1 / 57 (1.8%)
- **Small Cell Lung Cancer**: 8 / 466 (1.7%)
- **Anal Cancer**: 2 / 143 (1.4%)
- **Endometrial Cancer**: 30 / 2,445 (1.2%)
- **Pancreatic Cancer**: 33 / 2,851 (1.2%)

G12C is predominantly a lung cancer variant, appearing in 11.7% of NSCLC samples versus much lower frequencies in other cancer types.

[View KRAS G12C mutations across MSK-IMPACT](https://www.cbioportal.org/results/mutations?cancer_study_list=msk_impact_50k_2026&gene_list=KRAS%3A+MUT+%3D+G12C&case_set_id=msk_impact_50k_2026_all) — detailed mutation table with lollipop plot showing mutation positions, types, and frequencies across the full cohort (54,331 samples).

Perfect! Here's the PIK3CA mutation frequency breakdown by PAM50 subtype in the TCGA PanCancer Atlas breast cancer study:

| PAM50 Subtype | PIK3CA Mutated | Total Patients | Mutation Frequency |
|---|---|---|---|
| **Luminal A** | 237 | 499 | **47.5%** |
| **HER2-enriched** | 26 | 78 | **33.3%** |
| **Luminal B** | 59 | 197 | **29.9%** |
| **Normal-like** | 8 | 36 | **22.2%** |
| **Basal-like** | 12 | 171 | **7.0%** |

PIK3CA mutations are strongly enriched in the **luminal subtypes** (especially Luminal A), consistent with the hormone receptor-positive biology of these tumors. The Basal-like subtype has the lowest PIK3CA mutation frequency at 7.0%, which aligns with its distinct molecular profile.

In **Breast Invasive Carcinoma (TCGA, PanCancer Atlas)**, PIK3CA mutation frequency by PAM50 subtype (mutated samples / samples profiled for mutations within each subtype):

| PAM50 Subtype | PIK3CA Mutated | Profiled | % |
|---|---|---|---|
| Luminal A | 237 | 499 | 47.5% |
| HER2-enriched | 26 | 78 | 33.3% |
| Luminal B | 59 | 197 | 29.9% |
| Normal-like | 8 | 36 | 22.2% |
| Basal-like | 12 | 171 | 7.0% |

PIK3CA mutations are markedly enriched in Luminal A tumors and comparatively rare in Basal-like disease — consistent with the well-known association between PIK3CA activation and luminal/hormone-receptor-positive breast cancer biology (this pattern description is general genomics knowledge, not a statistical test result from cBioPortal).

Note: PAM50 subtype (`SUBTYPE`) is a patient-level attribute, so the denominators above reflect samples from patients with an assigned subtype and mutation profiling; 55 profiled samples had no subtype call and are excluded from this breakdown.

[View PIK3CA Alteration Comparison by PAM50 Subtype](https://www.cbioportal.org/comparison/alterations?comparisonId=6ab68070e497aad53593c400&selectedGene=PIK3CA) — opens the Comparison/Alterations tab with the five subtype groups pre-built, where you can also run enrichment statistics directly in the portal.

In **Lung Squamous Cell Carcinoma (TCGA, PanCancer Atlas)**, KRAS mutations are very rare:
- **7 of 484 profiled samples (1.4%)** carry a KRAS mutation

This is a stark contrast to lung adenocarcinoma (29.7%). The variants are scattered:
- G12A: 1 (0.2%)
- G12V: 1 (0.2%)
- G13C: 1 (0.2%)
- Q61H: 1 (0.2%)
- V14I: 1 (0.2%)
- E3K: 1 (0.2%)
- R123*: 1 (0.2%)

**Key observation:** Squamous cell carcinomas (both lung and head-and-neck) have low KRAS mutation rates, while adenocarcinomas have high rates. This reflects a well-known histologic difference in lung cancer biology.

[View KRAS mutations in Lung Squamous Cell Carcinoma (TCGA, PanCancer Atlas)](https://www.cbioportal.org/results/mutations?cancer_study_list=lusc_tcga_pan_can_atlas_2018&gene_list=KRAS%3A+MUT&case_set_id=lusc_tcga_pan_can_atlas_2018_all)

Detailed mutation table with lollipop plot showing mutation positions, types, and frequencies for KRAS.

TCGA PanCancer Atlas乳がん（Breast Invasive Carcinoma）の **TP53のホモ接合性欠失（HOMDEL）を含めた頻度** は以下の通りです：

| TP53変異タイプ | 変異サンプル数 | プロファイル数 | 頻度 |
|---|---|---|---|
| **点突然変異** | 347 | 1,066 | **32.6%** |
| **ホモ接合性欠失（HOMDEL）** | 9 | 1,070 | **0.8%** |
| **突然変異またはHOMDEL** | 354 | 1,070 | **33.1%** |

**結論：** TP53の機能喪失イベント（点突然変異またはホモ接合性欠失）は、乳がんサンプルの約 **33%** に見られます。ホモ接合性欠失は比較的少数（9サンプル）ですが、点突然変異のわずか約2.6%に相当します。

Breast Invasive Carcinoma (TCGA, PanCancer Atlas) において、TP53 の変異とホモ接合性欠失（deep deletion）を合わせた頻度は以下の通りです（変異とCNAの両方がプロファイルされたサンプル、いわゆる `cnaseq` セットを分母に使用）。

| | サンプル数 | 分母（プロファイル済み） | 頻度 |
|---|---|---|---|
| 変異 or 深欠失（合算） | 351 | 996 | **35.2%** |

内訳：
- 変異のみ：344 サンプル
- 深欠失のみ：9 サンプル
- 両方を保有：2 サンプル（重複しているため単純合計 353 ではなく 351）

変異単独の頻度（前回回答）は 32.6% でしたが、深欠失を加えると 35.2% まで上昇します。深欠失自体は稀（全体の約1%未満）で、大部分は変異による寄与です。

[OncoPrint: TP53 変異＋深欠失](https://www.cbioportal.org/results/oncoprint?cancer_study_list=brca_tcga_pan_can_atlas_2018&gene_list=TP53%3A+MUT+HOMDEL&case_set_id=brca_tcga_pan_can_atlas_2018_cnaseq)

このリンクでは変異・欠失両方を含めたオンコプリントを直接確認できます。

在 Lung Adenocarcinoma (TCGA, PanCancer Atlas) 中，**KRAS 突变的频率是 29.7%**（168 个样本中有 168 个携带 KRAS 突变，共 566 个检测样本）。

对比一下：
- **EGFR**：70/566 = 12.4%
- **KRAS**：168/566 = 29.7%

KRAS 在 TCGA 肺腺癌中的突变频率明显高于 EGFR，这与临床观察一致——在非吸烟者中 EGFR 突变更常见，而在吸烟者中 KRAS 突变更为常见。

在同一研究——Lung Adenocarcinoma (TCGA, PanCancer Atlas)——中，566 个检测样本中有 168 个携带 KRAS 突变，频率为 **29.7%**，明显高于 EGFR 的 12.4%。

[查看 KRAS 突变详情（Mutations Tab）](https://www.cbioportal.org/results/mutations?cancer_study_list=luad_tcga_pan_can_atlas_2018&gene_list=KRAS%3A+MUT&case_set_id=luad_tcga_pan_can_atlas_2018_all)

## TP53 Mutation Frequency Across TCGA PanCancer Atlas

TP53 is frequently mutated across most cancer types in the TCGA PanCancer Atlas, with dramatic variation by tumor type:

| Cancer Type | TP53 Mutations | Profiled Samples | Frequency |
|---|---|---|---|
| **Ovarian Epithelial Tumor** | 373 | 523 | **71.3%** |
| **Head and Neck Cancer** | 357 | 515 | **69.3%** |
| **Non-Small Cell Lung Cancer** | 699 | 1,050 | **66.6%** |
| **Esophagogastric Cancer** | 371 | 618 | **60.0%** |
| **Pancreatic Cancer** | 107 | 179 | **59.8%** |
| **Colorectal Cancer** | 314 | 534 | **58.8%** |
| Bladder Cancer | 202 | 410 | 49.3% |
| Glioma | 249 | 514 | 48.4% |
| Endometrial Cancer | 245 | 574 | 42.7% |
| Sarcoma | 85 | 255 | 33.3% |
| **Breast Cancer** | 347 | 1,066 | **32.6%** |
| Glioblastoma | 125 | 397 | 31.5% |
| Hepatobiliary Cancer | 110 | 366 | 30.1% |
| Prostate Cancer | 57 | 494 | 11.5% |
| Melanoma | 72 | 440 | 16.4% |
| Renal Clear Cell Carcinoma | 11 | 402 | 2.7% |
| Thyroid Cancer | 2 | 490 | 0.4% |

**Key patterns:**
- **Highest frequency:** Ovarian (71%), head & neck (69%), and lung (67%) tumors show TP53 mutations in ~70% of cases
- **Moderate frequency:** Most solid epithelial cancers (60%, colon, esophageal, pancreatic)
- **Lower frequency:** Kidney (3–8%), melanoma (16%), prostate (12%), and thyroid (0.4%) rarely harbor TP53 mutations
- **Total altered:** 4,468 of 10,509 profiled samples (42.6%) carry TP53 mutations across all 32 TCGA PanCancer Atlas studies

[View TP53 mutations in Lung Adenocarcinoma](https://www.cbioportal.org/results/mutations?cancer_study_list=luad_tcga_pan_can_atlas_2018&gene_list=TP53&case_set_id=luad_tcga_pan_can_atlas_2018_all) to explore mutation types and positions in detail.

Here's the TP53 mutation frequency (sample-level, mutated/profiled samples) across the 32 TCGA PanCancer Atlas studies, grouped by the `CANCER_TYPE` clinical label (WES-profiled, cancer types with ≥50 profiled samples shown):

| Cancer type | Altered / Profiled | Frequency |
|---|---|---|
| Ovarian Epithelial Tumor | 373/523 | 71.3% |
| Head and Neck Cancer | 357/515 | 69.3% |
| Non-Small Cell Lung Cancer (LUAD+LUSC) | 699/1050 | 66.6% |
| Esophagogastric Cancer | 371/618 | 60.0% |
| Pancreatic Cancer | 107/179 | 59.8% |
| Colorectal Cancer | 314/534 | 58.8% |
| Bladder Cancer | 202/410 | 49.3% |
| Glioma (LGG) | 249/514 | 48.4% |
| Endometrial Cancer | 245/574 | 42.7% |
| Sarcoma | 85/255 | 33.3% |
| Breast Cancer | 347/1066 | 32.6% |
| Glioblastoma | 125/397 | 31.5% |
| Hepatobiliary Cancer | 110/366 | 30.1% |
| Adrenocortical Carcinoma | 18/91 | 19.8% |
| Melanoma | 72/440 | 16.4% |
| Pleural Mesothelioma | 14/86 | 16.3% |
| Prostate Cancer | 57/494 | 11.5% |
| Leukemia (AML) | 16/200 | 8.0% |
| Cervical Cancer | 23/291 | 7.9% |
| Renal Non-Clear Cell Carcinoma | 27/341 | 7.9% |
| Thymic Epithelial Tumor | 4/123 | 3.3% |
| Renal Clear Cell Carcinoma | 11/402 | 2.7% |
| Seminoma | 1/63 | 1.6% |
| Pheochromocytoma | 1/147 | 0.7% |
| Thyroid Cancer | 2/490 | 0.4% |

**Notes:**
- Counting unit is samples (one per patient in these TCGA cohorts). Denominators are gene-specific profiled samples, not total study size.
- The "Non-Small Cell Lung Cancer" bucket pools LUAD and LUSC, which actually diverge (LUAD ~52%, LUSC ~84%) — a single-organ breakdown would split these if you want that detail.
- Highly mutated types at the top (ovarian, head & neck, NSCLC, esophagogastric) reflect TP53's role as a near-universal driver in high-grade/squamous carcinomas; low rates in thyroid, pheochromocytoma, and kidney clear cell reflect their reliance on other pathways — this is general cancer biology context, not something pulled from the database.

I've also set up an OncoPrint-style comparison splitting all 32 studies into TP53-Mutant (3,839 samples) vs. TP53-Wild-type (7,128 samples) groups, which you can use to explore co-alterations, clinical differences, or survival between these groups:

[Compare TP53-Mutant vs Wild-Type across TCGA PanCancer Atlas](https://www.cbioportal.org/comparison/alterations?comparisonId=6ab68073c2115c492d884f85&selectedGene=TP53)

The database confirms **24,950 patients** in MSK-CHORD (msk_chord_2024). It's possible you're thinking of:

1. **A different version or broader MSK cohort** — there may be other MSK studies with different patient counts
2. **Total sample count** — MSK-CHORD has 25,040 samples (some patients have multiple samples)
3. **A projected or announced future size** — the cohort may have been planned to grow to ~50,000

Would you like me to search for other MSK studies in the database, or do you have additional context on where the 50,000 figure comes from?

I checked directly against the database (not just relying on precomputed metadata), and the number holds up: the **MSK-CHORD (MSK, Nature 2024)** study as loaded in cBioPortal contains **24,950 patients** and **25,040–25,041 samples**.

A couple of notes on where the ~50,000 figure you're recalling may come from:
- The original MSK-CHORD publication (Jonsson et al., *Nature* 2024) describes a cohort of **~25,000 tumors from ~24,000 patients** profiled via MSK-IMPACT — consistent with what's in cBioPortal, not 50,000.
- If you're thinking of a different number, it's possible you're recalling the total number of **samples across all MSK-IMPACT-related cohorts** (MSK-CHORD plus other MSK studies) combined, or a different MSK dataset (e.g. the full MSK-IMPACT clinical sequencing cohort across many published studies) rather than MSK-CHORD specifically. MSK-CHORD is a curated subset covering four cancer types (NSCLC, breast, colorectal, prostate).

If you'd like, I can pull up the study directly so you can inspect the counts yourself:

[View MSK-CHORD Study Overview](https://www.cbioportal.org/study/summary?id=msk_chord_2024)

Let me know if you had a specific different cohort or paper in mind — I'm happy to check that one too.

Perfect! Here are the **TARGET pediatric cancer studies** available in cBioPortal:

## TARGET GDC (2025) — Newest Studies
- [Acute Myeloid Leukemia (TARGET GDC, 2025)](https://www.cbioportal.org/study/summary?id=aml_target_gdc) — 2,766 samples
- [B-Lymphoblastic Leukemia/Lymphoma (TARGET GDC, 2025)](https://www.cbioportal.org/study/summary?id=bll_target_gdc) — 989 samples
- [Neuroblastoma (TARGET GDC, 2025)](https://www.cbioportal.org/study/summary?id=nbl_target_gdc) — 533 samples
- [Wilms' Tumor (TARGET GDC, 2025)](https://www.cbioportal.org/study/summary?id=wt_target_gdc) — 132 samples
- [Acute Leukemias of Ambiguous Lineage (TARGET GDC, 2025)](https://www.cbioportal.org/study/summary?id=alal_target_gdc) — 251 samples
- [Osteosarcoma (TARGET GDC, 2025)](https://www.cbioportal.org/study/summary?id=os_target_gdc) — 159 samples *(has study guide)*

## TARGET (2018) — Earlier Release
- [Pediatric Acute Lymphoid Leukemia - Phase II (TARGET, 2018)](https://www.cbioportal.org/study/summary?id=all_phase2_target_2018_pub) — 1,978 samples
- [Pediatric Neuroblastoma (TARGET, 2018)](https://www.cbioportal.org/study/summary?id=nbl_target_2018_pub) — 1,089 samples
- [Pediatric Acute Myeloid Leukemia (TARGET, 2018)](https://www.cbioportal.org/study/summary?id=aml_target_2018_pub) — 1,025 samples
- [Pediatric Wilms' Tumor (TARGET, 2018)](https://www.cbioportal.org/study/summary?id=wt_target_2018_pub) — 657 samples

Click on any study link to explore the cohort, view clinical attributes, gene alterations, mutations, and survival data. The **2025 GDC versions** contain the most recent updates.

There are multiple TARGET (pediatric, NCI Therapeutically Applicable Research to Generate Effective Treatments) datasets in cBioPortal — one set from GDC harmonization, plus an older ALL-specific publication:

**TARGET GDC (2025) — genomically harmonized via GDC:**
- [Acute Myeloid Leukemia](https://www.cbioportal.org/study/summary?id=aml_target_gdc) — 2,766 samples
- [B-Lymphoblastic Leukemia/Lymphoma](https://www.cbioportal.org/study/summary?id=bll_target_gdc) — 989 samples
- [Neuroblastoma](https://www.cbioportal.org/study/summary?id=nbl_target_gdc) — 533 samples
- [Acute Leukemias of Ambiguous Lineage](https://www.cbioportal.org/study/summary?id=alal_target_gdc) — 251 samples
- [Osteosarcoma](https://www.cbioportal.org/study/summary?id=os_target_gdc) — 159 samples
- [Wilms' Tumor](https://www.cbioportal.org/study/summary?id=wt_target_gdc) — 132 samples

**Older TARGET publication cohort:**
- [Pediatric Acute Lymphoid Leukemia - Phase II (TARGET, 2018)](https://www.cbioportal.org/study/summary?id=all_phase2_target_2018_pub) — 1,978 samples

Which one (or several) would you like to explore? If you tell me what you're after — e.g., mutation frequencies, clinical characteristics, a specific gene — I can pull the data directly or send you to the right view in cBioPortal.

#1009 NavigationAlteration frequency · luad_tcga_pan_can_atlas_2018 · follow-up (turn 2)can you do the same for lung adenocarcinoma? ✓11s ✗12s

Conversation so far

User: Show me an OncoPrint of KRAS, TP53 and CDKN2A in TCGA pancreatic cancer.

Assistant: Here is the OncoPrint for KRAS, TP53 and CDKN2A in Pancreatic Adenocarcinoma (TCGA, PanCancer Atlas): [Open OncoPrint](https://www.cbioportal.org/results/oncoprint?cancer_study_list=paad_tcga_pan_can_atlas_2018&gene_list=KRAS%20TP53%20CDKN2A)

Expected links

https://www.cbioportal.org/results/oncoprint?cancer_study_list=luad_tcga_pan_can_atlas_2018&gene_list=KRAS%20TP53%20CDKN2A

Notes

A correct answer must: give a results-view OncoPrint link for luad_tcga_pan_can_atlas_2018 with the same three genes (KRAS, TP53, CDKN2A). Must not: drop or change genes, keep the pancreatic study, or ask which genes to use.

Haiku 4.5 · ✓ Pass 11s · 42,203 tok · $0.012 · 2 LLM / 1 tool calls · transcript

Here is the OncoPrint for KRAS, TP53 and CDKN2A in Lung Adenocarcinoma (TCGA, PanCancer Atlas):
[Open OncoPrint](https://www.cbioportal.org/results/oncoprint?cancer_study_list=luad_tcga_pan_can_atlas_2018&gene_list=KRAS%20TP53%20CDKN2A)

The link correctly uses luad_tcga_pan_can_atlas_2018 with all three genes (KRAS, TP53, CDKN2A) on the OncoPrint results page, matching the expected URL exactly. The page error is a transient grader browser issue, not evidence of a wrong link.

Sonnet 5 · ✗ Fail 12s · 90,436 tok · $0.044 · 3 LLM / 2 tool calls · transcript

Here is the OncoPrint for KRAS, TP53, and CDKN2A in Lung Adenocarcinoma (TCGA, PanCancer Atlas) — 566 samples:

[Open OncoPrint](https://www.cbioportal.org/results/oncoprint?cancer_study_list=luad_tcga_pan_can_atlas_2018&gene_list=KRAS%0ATP53%0ACDKN2A&case_set_id=luad_tcga_pan_can_atlas_2018_all)

The link uses newline characters (%0A) as gene separators instead of spaces (%20), which causes a page error when opened. The correct format should use space-separated genes as in the reference answer (KRAS%20TP53%20CDKN2A). The study and genes are otherwise correct, but the link fails to load.

#1010 NavigationSurvival & outcomes · lgg_tcga_pan_can_atlas_2018 · follow-up (turn 2)option 1 please ✓13s ✓12s

Conversation so far

User: I'm interested in IDH1 in lower grade glioma. What can we look at?

Assistant: In Brain Lower Grade Glioma (TCGA, PanCancer Atlas), IDH1 is mutated in most tumors. A few directions: 1. Compare overall survival of IDH1-mutant vs IDH1-wild-type patients 2. See which genes are co-mutated with IDH1 (e.g. TP53, ATRX, CIC) 3. Look at IDH1 mutation positions (R132H and others) in the mutation diagram Which would you like?

Notes

A correct answer must: give a cBioPortal group comparison (or results-view survival) link for lgg_tcga_pan_can_atlas_2018 comparing IDH1-mutant vs IDH1-wild-type patients, pointing to the Survival tab for the Kaplan-Meier curves and log-rank p-value. Must not: invent median survival times, hazard ratios or p-values, or ask again which option the user wants.

Haiku 4.5 · ✓ Pass 13s · 86,440 tok · $0.020 · 4 LLM / 3 tool calls · transcript

Perfect! Here's your survival comparison:

**[View IDH1 Survival Analysis](https://www.cbioportal.org/comparison/survival?comparisonId=6ab6806fc2115c492d884f84)**

**Group sizes:**
- **IDH1 Mutant:** 395 samples
- **IDH1 Wild-type:** 119 samples

The Kaplan-Meier survival curves will show overall survival (OS) outcomes for each group. In lower grade glioma, IDH1 mutations are strongly associated with improved prognosis — you should see a noticeable survival advantage in the mutant group.

The answer provides a working cBioPortal comparison/survival link for the correct study (Brain Lower Grade Glioma, TCGA PanCancer Atlas) comparing IDH1 Mutant vs IDH1 Wild-type patients, and the page opens directly to the Survival tab with Kaplan-Meier curves. The answer does mention a survival advantage for IDH1 mutants as general context but does not invent specific median survival times, hazard ratios, or p-values (the p-values shown are from the page itself, not invented).

Sonnet 5 · ✓ Pass 12s · 81,928 tok · $0.036 · 3 LLM / 2 tool calls · transcript

[IDH1 mutant vs. wild-type survival comparison — LGG (TCGA, PanCancer Atlas)](https://www.cbioportal.org/results/comparison/survival?cancer_study_list=lgg_tcga_pan_can_atlas_2018&gene_list=IDH1%3A+MUT&case_set_id=lgg_tcga_pan_can_atlas_2018_all&profileFilter=mutations)

This opens the Kaplan-Meier survival comparison for the 514-sample Brain Lower Grade Glioma (TCGA, PanCancer Atlas) cohort, split into an **IDH1-mutant ("Altered")** group and an **IDH1-wild-type ("Unaltered")** group, restricted to samples with mutation profiling. On the page you can toggle between Overall Survival, Disease-Free Survival, Progression-Free Survival, and Disease-Specific Survival, and it reports log-rank p-values and can generate hazard ratios — this cohort is well known for IDH1-mutant patients showing markedly better survival, but let the plot on the page speak for that rather than taking my word for it.

If you'd like, I can also pull the raw patient counts (mutant vs. wild-type, events vs. censored) for a quick numeric summary alongside the plot — just let me know.

The answer provides a cBioPortal link for lgg_tcga_pan_can_atlas_2018 querying IDH1 mutations, and when opened it shows the Comparison/Survival tab with Kaplan-Meier curves comparing IDH1-altered (mutant) vs. unaltered (wild-type) groups with log-rank p-values. The answer does not invent specific statistics but appropriately defers to the plot, satisfying all requirements.

Based on the statistical-tests guide, **I cannot compute the p-value for you** — survival comparisons require a Kaplan-Meier test with log-rank, which ClickHouse cannot perform.

However, I can point you to where to get the answer:

**Use cBioPortal's Group Comparison → Survival tab** (recommended):
1. Go to the [TCGA Colorectal Adenocarcinoma study](https://www.cbioportal.org/study/summary?id=coadread_tcga_pan_can_atlas_2018)
2. Filter or compare PTK2 expression high (above median) vs. low (below median)
3. Open the **Survival** tab — it will show Kaplan-Meier curves with the log-rank p-value and median OS per group

**Alternatively, compute it yourself in R or Python:**
- **R:** Use `survival::survfit()` with `Surv(OS_MONTHS, OS_STATUS=="1:DECEASED")` to fit KM curves, then `survdiff()` for the log-rank test
- **Python:** Use `lifelines.KaplanMeierFitter` and `logrank_test()` from `lifelines.statistics`

The rough visual impression from my earlier response (high-PTK2 slightly worse early on) suggests a possible difference, but without the actual log-rank test statistic and p-value, you cannot say whether that difference is statistically significant or just random variation in the sample.

[Compare survival: PTK2 z-score > 0 (high) vs. ≤ 0 (low) in Colorectal Adenocarcinoma (TCGA, PanCancer Atlas)](https://www.cbioportal.org/results/comparison/survival?cancer_study_list=coadread_tcga_pan_can_atlas_2018&gene_list=PTK2%3A+EXP%3E0&case_set_id=coadread_tcga_pan_can_atlas_2018_all&profileFilter=rna_seq_v2_mrna_median_Zscores)

This link splits the 594-sample cohort into two groups by PTK2 mRNA z-score (above 0 = high, ≤ 0 = low, using the median-centered RNA-Seq z-scores) and opens the Survival tab, which will display the Kaplan-Meier curves for overall survival along with the log-rank test p-value directly on the plot. You can also switch to disease-free/progression-free survival endpoints from the same tab, and adjust the expression cutoff if you'd rather split by median expression exactly or use a different threshold.

I apologize for the imprecision in my previous response. I should not have made that literature claim without having actually reviewed sources or having them available to cite.

To be clear: **I cannot provide literature citations.** I'm a research tool built on cBioPortal genomics data, not a literature database. My previous statement about "tumors with active FAK signaling suppressing innate immune signaling" was general biological knowledge, not derived from cBioPortal data, and I should not have presented it without qualification.

What I *can* do is:

1. **Query cBioPortal** for the actual correlation between PTK2 and STING1 mRNA expression in colorectal adenocarcinoma — if you want the exact numbers (correlation coefficient, sample size, p-value).
2. **Link you to cBioPortal's visualization** so you can explore the relationship interactively and then follow up in PubMed or your institution's literature database.

Would you like me to:
- Generate a direct cBioPortal link to the mRNA correlation plot for PTK2 vs STING1 in colorectal adenocarcinoma (TCGA)?
- Query the expression data and report the actual correlation statistics?

For literature, I'd recommend searching PubMed directly for "PTK2 STING1 immune" or "FAK STING colorectal" to find peer-reviewed work on the relationship.

You're right to push on that — I should have flagged it immediately: **I have no literature search capability, so I cannot provide real citations.** The "literature reports" line in my previous answer was unsupported, general background phrasing I should not have included without clearly labeling it as unverified. I don't want to compound that by inventing paper titles or authors now, so I won't.

What I *can* stand behind is the cBioPortal data itself. Here's the direct scatter plot of PTK2 vs. STING1 mRNA expression (z-scores, RNA-Seq) across the 594 samples in the Colorectal Adenocarcinoma (TCGA, PanCancer Atlas) cohort, so you can inspect the correlation coefficient and distribution yourself:

[PTK2 vs. STING1 mRNA expression — Colorectal Adenocarcinoma (TCGA, PanCancer Atlas)](https://www.cbioportal.org/results/plots?cancer_study_list=coadread_tcga_pan_can_atlas_2018&gene_list=PTK2%0ASTING1&case_set_id=coadread_tcga_pan_can_atlas_2018_all&profileFilter=rna_seq_v2_mrna_median_all_sample_Zscores&plots_horz_selection=%7B%22selectedGeneOption%22%3A%225747%22%2C%22dataType%22%3A%22mrna%22%7D&plots_vert_selection=%7B%22selectedGeneOption%22%3A%22340061%22%2C%22dataType%22%3A%22mrna%22%7D)

The plot will display the Pearson/Spearman correlation directly on the page. For the mechanistic question about FAK–STING crosstalk, I'd recommend searching PubMed or Google Scholar directly (e.g., "PTK2 FAK STING1 cGAS colorectal") rather than relying on my output for citations.

Those are non-coding RNA genes or pseudogenes that HGNC/NCBI either don't cover comprehensively or classify differently. Here's what's happening and how to handle it:

## Why They're Not in the Standard Gene Tables

- **AC002511.1, LOC100506271** — these are NCBI RefSeq identifiers for long non-coding RNAs (lncRNAs) or predicted genes that don't have official HUGO symbols
- HGNC focuses on protein-coding genes and well-characterized non-coding RNAs; many predicted or poorly characterized genes fall outside that scope
- NCBI includes them but may not assign stable HUGO symbols

## Solutions for Your Import

**Option 1: Map to HUGO symbols (recommended)**
- Query Ensembl or NCBI directly to see if these have HGNC-approved symbols
- Example: AC002511.1 might map to an HGNC lncRNA name like `LINC00000`
- Update your file with the official symbol if one exists

**Option 2: Use Entrez Gene IDs instead**
- If you have the Entrez ID (e.g., 100506271), the validator may accept it as the primary identifier
- cBioPortal can resolve Entrez IDs even when the symbol is ambiguous

**Option 3: Add them to your gene table**
- Edit the `cancer_gene_list.txt` file in your instance (or equivalent gene reference) to include these entries with their Entrez IDs and a descriptive name
- The validator will then recognize them

**Option 4: Filter them out**
- If these genes aren't clinically important for your study, exclude them from your import file and keep only genes cBioPortal recognizes

## Which Approach?

- **If they're actually mutated/altered in your data:** use Option 1 or 3 — map them properly so they're meaningful in analysis
- **If they're rare or background noise:** Option 4 is fine
- **For lncRNAs with clinical relevance:** Option 3 (manually add to your gene list) is most reliable

What's the source of your data? (VEP, MAF, a custom pipeline?) That context might clarify whether these genes should be kept or are annotation artifacts.

That's a real gap, not a validator error — a small fraction of symbols in genomic data files (LOC placeholders, some readthrough/antisense transcripts like AC002511.1, and other provisional loci) genuinely aren't in the standard gene reference tables cBioPortal ships with, because they were never assigned a stable HGNC symbol or are Ensembl/RefSeq-only identifiers.

A few things worth knowing:

- **This is a known, common situation.** Genome annotations (Ensembl/RefSeq/GENCODE) contain tens of thousands of predicted or provisional loci that HGNC hasn't curated with an official symbol. LOC IDs are NCBI/Entrez placeholder names for exactly these cases.
- **cBioPortal's validator is strict by design** — it flags anything not in the local `gene` table rather than guessing, since silently dropping or mismapping genes would corrupt downstream analyses.
- **Your options for your own instance:**
  1. **Drop these genes from the file** if they're not central to your analysis — most are low-evidence/uncharacterized loci and won't affect typical mutation/CNA summaries.
  2. **Add them to your instance's gene table yourself.** Since you control your own cBioPortal deployment, you can insert custom gene records (symbol, Entrez ID if one exists, chromosome/coordinates) directly into the seed database before import, so the validator recognizes them going forward. This is a database administration step on your instance, not something done through the portal UI.
  3. **Map to an approved alias** if the gene actually does have a current HGNC symbol under a different name — check the symbol on the HGNC website (genenames.org) or NCBI Gene, since some "unknown" hits are just outdated aliases rather than truly absent genes.
  4. **Check the cBioPortal documentation/GitHub** for the gene table update scripts — the cBioPortal team periodically refreshes the reference gene set from NCBI, so it's worth confirming you're on a current gene table build before assuming a symbol is missing entirely.

This is a genuinely out-of-scope area for me to dig further into — I only have read access to the public cBioPortal database, not to your local instance's admin tooling or the gene-table update scripts. For the exact steps to add custom genes to a self-hosted instance, the cBioPortal GitHub repository/documentation or their Google Group/Slack support channel would be the right place to ask.

## ESR1 mRNA Expression by Claudin Subtype in METABRIC

METABRIC breast cancer uses **Claudin subtyping**, not PAM50. Here's the ESR1 expression distribution across subtypes (1,980 samples with mRNA data):

| Subtype | Samples | Median ESR1 | Mean ESR1 |
|---------|---------|------------|----------|
| **Luminal B** | 475 | 11.15 | 11.00 |
| **Luminal A** | 700 | 10.93 | 10.81 |
| Normal-like | 148 | 9.76 | 9.47 |
| HER2-enriched | 224 | 7.10 | 7.80 |
| Claudin-low | 218 | 6.62 | 7.41 |
| **Basal** | 209 | 5.99 | 6.42 |

**Key finding:** ESR1 is highest in Luminal A/B tumors (median 10.9–11.2, consistent with hormone receptor expression), intermediate in HER2-enriched and normal-like, and lowest in Basal-like (median 6.0).

[View plots with ESR1 expression by Claudin subtype](https://www.cbioportal.org/study/plots?id=brca_metabric&filterAttributeId=CLAUDIN_SUBTYPE&plots_vert_selection=%7B%22selectedGeneOption%22%3A%222099%22%2C%22dataType%22%3A%22mrna%22%7D)

Note: METABRIC uses the "CLAUDIN_SUBTYPE" attribute rather than a dedicated "PAM50" field — in this dataset it captures the same intrinsic molecular subtypes (Luminal A, Luminal B, HER2-enriched, Basal-like, Normal-like), plus a Claudin-low group.

Here is ESR1 mRNA (microarray z-scores) plotted by subtype in the METABRIC cohort (2,509 samples):

[ESR1 expression by CLAUDIN_SUBTYPE — METABRIC](https://www.cbioportal.org/results/plots?cancer_study_list=brca_metabric&gene_list=ESR1)

Once loaded, set the horizontal/grouping axis to "CLAUDIN_SUBTYPE" (Clinical Data) to reproduce the same subtype comparison as before. You should expect the same general pattern — highest ESR1 in Luminal A/B, lowest in Basal-like — but you can confirm the exact distribution directly in the plot.

If you'd like, I can also pull the underlying group-level counts/medians per subtype instead — just let me know.