[cbhindex]

Until now the neuroblastoma entry recorded telomere maintenance only in prose: one sentence in the description and two genetic records (TERT Rearrangement and ATRX Mutation, from #7114). Nothing in the pathophysiology section modelled it. This PR adds that mechanism layer: four nodes and four causal edges, with 54 evidence items from 17 papers, 15 of them new to the entry. It also replaces the unsourced "treatment-refractory" wording about ATRX-altered disease with what the sources actually report, and swaps a bioRxiv preprint in references: for its published version. Part of #14976.

graph LR
  MYCN["MYCN-Driven Proliferation (existing)"] -->|raises TERT transcription| TA[Telomerase Activation]
  TA --> URP[Unlimited Replicative Potential]
  AD[Reduced ATRX/DAXX Complex Abundance] -->|permits| ALT[Alternative Lengthening of Telomeres]
  ALT --> URP

Loading

Nodes

Node	Bindings	conforms_to	Main content
Telomerase Activation	TERT; MF GO:0003720 telomerase activity (increased); BP GO:0007004	enabling_replicative_immortality#Telomere Maintenance Reactivation	the routes (TERT rearrangement with enhancer juxtaposition, MYCN amplification, high TERT expression without either, rare promoter mutations); frequencies; activated telomerase as an independent marker of poor overall survival
Reduced ATRX/DAXX Complex Abundance	ATRX; MF GO:0140658 ATP-dependent chromatin remodeler activity (decreased)	none	ATRX/DAXX function; reduced complex in ALT tumours, by mutation or low protein; lesion spectrum (mostly multi-exon deletions, often in-frame); age distribution
Alternative Lengthening of Telomeres	BP GO:0000722	none (see below)	frequency by assay and cohort, including relapse; ATRX alterations in only about half to 60%; near-exclusivity with telomerase routes; slow growth but poor outcome, including in tumours not classed high risk
Unlimited Replicative Potential	BP GO:0008283	#Replicative Immortality	consequence node: replicative lifespan rather than faster growth; when telomere maintenance is acquired; outcome by telomere maintenance status

Design choices

• Two mechanisms, not three routes. Telomerase activation and ALT are different mechanisms with different GO processes and different module anchors, so they get separate nodes. MYCN amplification is a route into telomerase activation, because amplified MYCN raises TERT transcription. It is therefore an edge from the existing MYCN node, not a node of its own. The edge is INDIRECT_KNOWN_INTERMEDIATES, with the intermediate named, because its source node is named for proliferation.
• ATRX loss is its own node. ATRX alterations explain only about half to 60% of ALT tumours, so ALT cannot carry ATRX as its defining gene.
• Gene-activity grounding. Putting TERT and ATRX on nodes through genes: brings both under check_gene_activity_grounding. Each lands on a node with a molecular function, and the check passes against origin/main.
• Counter-evidence is recorded as REFUTE.
  • Dagg 2017 (PMID:28636942) on Unlimited Replicative Potential: a high-risk subset lacks telomere lengthening activity, and cells derived from it grew for more than 200 population doublings while their telomeres shortened.
  • Koneru 2020 (PMID:32291317) on the MYCN edge: some MYCN-amplified tumours have TERT expression as low as ALT tumours, and MYCN overexpression did not overcome TERT repression in cell lines.
• Consequence-node binding. GO:0008283 (increased) copies the anchor #Replicative Immortality, as Glioblastoma_IDH_Wildtype does. The description says it means growth continuing past the telomere limit, not faster division, because ALT tumours grow more slowly than telomerase-activated ones.

Genetic records and description

• ATRX Mutation. The notes no longer say "treatment-refractory" or "essentially absent in infants". They now say:

  • poor response to induction chemotherapy is reported for large ATRX deletions and truncating mutations in one two-centre retrospective cohort (Sun 2026, PMID:42558321);
  • chemo-resistance with late recurrence is reported in a small series (Kurihara 2014, PMID:25487495);
  • ATRX alteration is rare in infancy, not absent.

  Six evidence items were added. One existing explanation was corrected: whole-genome sequencing covered 40 of Cheung's 104 patients, not all 104.

• TERT Rearrangement. The notes and one explanation said "three alternative routes to telomere maintenance". They now use the two-mechanism framing, with TERT rearrangement as a route into telomerase activation.

• Description. The telomere sentence now names the two mechanisms. It defines the third telomerase route as high TERT expression without either alteration, and limits the chemotherapy finding to large ATRX deletions and truncating mutations.

• References. The preprint DOI:10.1101/2022.08.16.504100 is replaced in place by its published version, PMID:39635126 (same title). Its abstract is also cited as evidence on the ALT node. The preprint's cache file stays in the repository; nothing cites it any more.

Points a reviewer may question

• No conforms_to on the ALT node. The right anchor, #Telomere Maintenance via Recombination, exists only in open maintainer PR Flesh out mechanism modules: 5 new, 21 extended, ~100 conformance wirings, 3 new disorder entries, and a conformance-content check #13924. The conforms_to foreign-key check runs only in the merge queue, so pointing at it now would pass here and be ejected later. Conforming to the telomerase anchor instead would claim a mechanism ALT tumours do not use. Once Flesh out mechanism modules: 5 new, 21 extended, ~100 conformance wirings, 3 new disorder entries, and a conformance-content check #13924 merges, a one-line follow-up adds the conformance; Leiomyosarcoma's ALT node, which Flesh out mechanism modules: 5 new, 21 extended, ~100 conformance wirings, 3 new disorder entries, and a conformance-content check #13924 leaves on the telomerase anchor, should move in the same follow-up.
• No separate TERT enhancer-hijacking node. The rearrangement itself stays in the TERT Rearrangement record, which now links to Telomerase Activation through genes:. The node description and evidence cover the enhancer juxtaposition.
• One paper, two evidence sources. Peifer 2015, Koneru 2020 and Dagg 2017 have both patient-tumour sentences (HUMAN_CLINICAL) and cell-line sentences (IN_VITRO). Each item is graded by its own sentence, as CLAUDE.md asks, and check-snippet-grading passes.
• Background quotes for ATRX's normal function. Two BACKGROUND items (Cheung 2012, van Gerven 2022) support the chromatin-remodeller molecular function. They are graded IN_VITRO, after the biochemical and cell work they restate.
• Outcome evidence on the consequence node is INDIRECT. A survival association does not measure replicative capacity, and every explanation on that node says so on the same basis.
• Where the MYCN edge sits. It is placed between the MYCN node's biological_processes and evidence blocks, because Neuroblastoma: wire ALK and MYCN to their mechanism nodes #15084 appends evidence at the end of that node. git merge-tree against Neuroblastoma: wire ALK and MYCN to their mechanism nodes #15084, and against the other open neuroblastoma PRs, shows no conflicts and no duplicate keys.
• Relation to Pediatric cancer review + fixes to 7 entries (WHO 2021 MB, INSS 4S, Osteosarcoma histopath/biochem) #7114. The pediatric-cancer review report marks the neuroblastoma telomere axis as done under Pediatric cancer review + fixes to 7 entries (WHO 2021 MB, INSS 4S, Osteosarcoma histopath/biochem) #7114. This PR adds the mechanism layer on top of those genetic records; it does not reopen them.
• Size. 54 items is a lot for one PR. They cover one mechanism, and most nodes need cohort-specific figures, because frequencies and outcomes differ a great deal by assay and risk group.

Not in this PR

• No variant_origin on any new node, so the entry stays NO_ORIGIN for check-cancer-origin. Telomere maintenance is acquired during tumour evolution (Körber 2023), not the transforming event.
• No telomere-directed or ATM-directed treatments; that evidence is preclinical.
• Phenotypes are not yet wired to these nodes.
• Ackermann 2018 (PMID:30523111) is cited only for telomere maintenance versus prognosis. Later PRs on relapse pathways and regression will reuse its cache.
• Seven references fetched while checking sources are not cited and not committed: PMID:11074604, 20006006, 20651253, 26523776, 38136279, 38837897, 42848963.

Validation

• just validate; just validate-terms (every new CURIE was already in the term and enum caches).
• just count-verified-snippets: 106/106. Ten of the new caches hold full text, and several snippets match text beyond the abstract.
• just validate-disorders: 106 snippets, 114 titles, 0 issues.
• scripts/check_gene_activity_grounding.py --against-ref origin/main: no newly ungrounded genes.
• python -m dismech.node_classes --verify-kb: all examples resolve; MYCN-Driven Proliferation is unchanged.
• Duplicate-key, entity-ref, causal-target, block-scalar, enum-value and folded-hyphen checks, plus the whole-KB snippet-length, title-snippet, snippet-grading, retired-prose, reference-title, case-collision, term-cache-integrity, shared-disease-term and coarse-phenotype gates.
• just prose-figure-audit: no flags for the new prose. Its two UNCITED rows are older figures in the description and in a subtype that Neuroblastoma: INRGSS stages, high-risk and adrenal subtypes #15100 removes.
• just check-cancer-origin: still NO_ORIGIN, as intended.
• The page renders, showing the four nodes.

The PR adds 15 reference caches and a history record.

🤖 Generated with Claude Code

[ai4c-reviewer]

Claude finished @cbhindex's task in 1m 53s —— View job

---

Review: Neuroblastoma telomere maintenance mechanisms

• Read the diff and check scope
• Run validation and snippet checks
• Review biology, ontology terms and evidence
• Post review (APPROVED)

Approving. The new mechanism layer is accurate and atomic, the evidence is real and quoted exactly, and the edits to the older notes fix claims the sources did not support. I found nothing blocking. Three optional suggestions are at the end.

• Schema validation would pass (just validate: passed)
• All PMIDs verified as real (15 new caches; I checked all 15 cache titles against the reference_title strings and every one matches)
• Evidence snippets are exact quotes (just count-verified-snippets: 106/106)
• Ontology terms correct and specific (just validate-terms: passed)
• evidence_source classifications correct
• No pathophysiology bundling (atomic entries only)

What I checked

• Scope. The PR touches 17 files: the entry, 15 reference caches and one history record. All of them belong to this change, and I saw no reverted or unrelated churn.
• Graph. The MYCN → Telomerase Activation → Unlimited Replicative Potential chain and the ATRX/DAXX loss → ALT → Unlimited Replicative Potential chain resolve. check-causal-targets reports no new dangling targets and check-entity-refs passes. Both conforms_to anchors exist in enabling_replicative_immortality.yaml.
• Ontology placement. Each GO term sits in the right slot. GO:0003720 and GO:0140658 are molecular functions and are under molecular_functions. GO:0007004, GO:0000722 and GO:0008283 are biological processes and are under biological_processes. Telomerase activation and ALT are separate nodes because they are different mechanisms, and ATRX/DAXX loss has its own node because ATRX alterations explain only about half to 60% of ALT tumours. Both choices are correct.
• Evidence grading.
  • Where one paper reports both patient tumours and cell lines (Peifer 2015, Koneru 2020, Dagg 2017), each item is graded by its own sentence.
  • quote_role: BACKGROUND is used correctly for introduction and discussion sentences.
  • Survival associations on the consequence node are marked INDIRECT, which is correct.
  • The REFUTE items are well chosen and stated precisely: Koneru on the MYCN edge, and Dagg's high-risk tumours that lack a telomere maintenance mechanism.
• Corrected claims. "Treatment-refractory" and "essentially absent in infants" were not supported by the cited sources. They are now replaced with what the sources show, each with a citation: Sun 2026 for poor induction response, limited to large deletions and truncating mutations; Kurihara 2014 for chemo-resistance with late recurrence; and the 3 of 81 infant cases.
• Preprint swap. Nothing in the entry still cites the preprint DOI, and the published PMID:39635126 has the same title.
• Deep research and GeneReviews. This is an augmentation of a somatic cancer entry. No new deep-research artifact is expected, and the GeneReviews baseline does not apply.

🔵 Suggestions (optional)

1. Unlimited Replicative Potential binding. GO:0008283 cell population proliferation with INCREASED copies the module anchor. However, the node's own description says it does not mean faster division, so the binding and the description pull in different directions. Following the module is defensible, so keep it if you prefer. It would be worth a sentence in the module's own description at some point, because every conformer inherits the same tension.
2. ALT → Unlimited Replicative Potential evidence, PMID:40638875 item. This BACKGROUND snippet is about telomere maintenance mechanisms in general rather than ALT, and the PMID:24792489 item already makes the same point more specifically. You could drop it with no loss.
3. Follow-ups you have already listed. The ALT conforms_to once Flesh out mechanism modules: 5 new, 21 extended, ~100 conformance wirings, 3 new disorder entries, and a conformance-content check #13924 merges, and wiring phenotypes to the new nodes. Both look right, and I agree with deferring them.

[ai4c-reviewer]

Review: Neuroblastoma telomere maintenance mechanisms

Approving. The new mechanism layer is accurate and atomic, the evidence is real and quoted exactly, and the edits to the older notes fix claims the sources did not support. I found nothing blocking. Three optional suggestions are at the end.

• Schema validation would pass (just validate: passed)
• All PMIDs verified as real (15 new caches; I checked all 15 cache titles against the reference_title strings and every one matches)
• Evidence snippets are exact quotes (just count-verified-snippets: 106/106)
• Ontology terms correct and specific (just validate-terms: passed)
• evidence_source classifications correct
• No pathophysiology bundling (atomic entries only)

What I checked

• Scope. The PR touches 17 files: the entry, 15 reference caches and one history record. All of them belong to this change, and I saw no reverted or unrelated churn.
• Graph. The MYCN → Telomerase Activation → Unlimited Replicative Potential chain and the ATRX/DAXX loss → ALT → Unlimited Replicative Potential chain resolve. check-causal-targets reports no new dangling targets and check-entity-refs passes. Both conforms_to anchors exist in enabling_replicative_immortality.yaml.
• Ontology placement. Each GO term sits in the right slot. GO:0003720 and GO:0140658 are molecular functions and are under molecular_functions. GO:0007004, GO:0000722 and GO:0008283 are biological processes and are under biological_processes. Telomerase activation and ALT are separate nodes because they are different mechanisms, and ATRX/DAXX loss has its own node because ATRX alterations explain only about half to 60% of ALT tumours. Both choices are correct.
• Evidence grading.
  • Where one paper reports both patient tumours and cell lines (Peifer 2015, Koneru 2020, Dagg 2017), each item is graded by its own sentence, as CLAUDE.md asks.
  • quote_role: BACKGROUND is used correctly for introduction and discussion sentences (Cheung 2012, van Gerven 2022, the 39635126 framing sentence).
  • Survival associations on the consequence node are marked INDIRECT, which is correct.
  • The REFUTE items are well chosen and stated precisely: Koneru on the MYCN edge, and Dagg's high-risk tumours that lack a telomere maintenance mechanism.
• Corrected claims. "Treatment-refractory" and "essentially absent in infants" were not supported by the cited sources. They are now replaced with what the sources show, each with a citation: Sun 2026 for poor induction response, limited to large deletions and truncating mutations; Kurihara 2014 for chemo-resistance with late recurrence; and the 3 of 81 infant cases.
• Preprint swap. Nothing in the entry still cites the preprint DOI (0 hits), and the published PMID:39635126 has the same title.
• Deep research and GeneReviews. This is an augmentation of a somatic cancer entry. No new deep-research artifact is expected, and the GeneReviews baseline does not apply.

🔵 Suggestions (optional)

1. Unlimited Replicative Potential binding. GO:0008283 cell population proliferation with INCREASED copies the module anchor. However, the node's own description says it does not mean faster division, so the binding and the description pull in different directions. Following the module is defensible, so keep it if you prefer. It would be worth a sentence in the module's own description at some point, because every conformer inherits the same tension.
2. ALT → Unlimited Replicative Potential evidence, PMID:40638875 item. This BACKGROUND snippet is about telomere maintenance mechanisms in general rather than ALT, and the PMID:24792489 item already makes the same point more specifically. You could drop it with no loss.
3. Follow-ups you have already listed. The ALT conforms_to once #13924 merges, and wiring phenotypes to the new nodes. Both look right, and I agree with deferring them.

[cbhindex]

Thanks for the review. Replies to the three suggestions, none of which changes this PR:

1. Unlimited Replicative Potential binding. I'm keeping GO:0008283 (increased) so the node matches its anchor, enabling_replicative_immortality#Replicative Immortality, as Glioblastoma_IDH_Wildtype does. I agree the tension belongs to the module, not to this entry: a sentence in the module's description saying the binding means continued expansion past the telomere limit, not a faster division rate, would cover every conformer. That is a module edit, and the module is being reworked in Flesh out mechanism modules: 5 new, 21 extended, ~100 conformance wirings, 3 new disorder entries, and a conformance-content check #13924, so I'll raise it there rather than here.
2. PMID:40638875 item on the ALT → Unlimited Replicative Potential edge. Agreed, it is the weaker of the two background items. I'll drop it in the follow-up that adds the ALT node's conforms_to once Flesh out mechanism modules: 5 new, 21 extended, ~100 conformance wirings, 3 new disorder entries, and a conformance-content check #13924 merges, since that PR edits the same node anyway.
3. Follow-ups. Yes: the ALT conformance (with Leiomyosarcoma's ALT node re-anchored at the same time) after Flesh out mechanism modules: 5 new, 21 extended, ~100 conformance wirings, 3 new disorder entries, and a conformance-content check #13924, and wiring phenotypes to these nodes in the phenotype-wiring PR.

Unassigning myself now so the auto-merge controller can take it after its 3-day window.