SNF8-Related Neurodevelopmental Disorder

1. Disease Information

Overview

SNF8-related neurodevelopmental disorder is an autosomal recessive allelic spectrum caused by bi-allelic loss-of-function variants in SNF8, which encodes one of three subunits of the ESCRT-II complex (Endosomal Sorting Complex Required for Transport II). The disorder was delineated in 2024 and spans a strikingly wide severity range that OMIM has split into two separate phenotype entries:

The verbatim framing from the defining paper (PMID:38423010, quotable snippet):

"We report nine individuals from six families presenting with a spectrum of neurodevelopmental/neurodegenerative features caused by bi-allelic variants in SNF8 (GenBank: NM_007241.4), encoding the ESCRT-II subunit SNF8."

"The phenotypic spectrum included four individuals with severe developmental and epileptic encephalopathy, massive reduction of white matter, hypo-/aplasia of the corpus callosum, neurodevelopmental arrest, and early death. A second cohort shows a milder phenotype with intellectual disability, childhood-onset optic atrophy, or ataxia."

Key identifiers

Synonyms / alternative names

• SNF8-related neurodevelopmental disorder (umbrella term; used as this entry's name)
• Developmental and epileptic encephalopathy 115 / DEE115 (severe pole)
• Neurodevelopmental disorder plus optic atrophy / NEDOA (mild pole)
• SNF8-related ESCRT-II deficiency
• Gene-level aliases relevant to literature searching: EAP30, VPS22, Dot3 (searching "VPS22" and "EAP30" retrieves the cell-biology literature that does not use the symbol SNF8)

Evidence provenance type

Aggregated disease-level, case-series derived. All content originates from deep-phenotyped individual case descriptions in a single multi-center research collaboration (Munich, Leipzig, Rome, Bologna, Jerusalem, Ulm, Oslo, Nürnberg), plus structured secondary curation in OMIM, ClinVar, MONDO, and Genomics England PanelApp. No EHR-derived, registry, or population-cohort data exist.

2. Etiology

Disease causal factors

Monogenic, fully genetic. The sole established cause is bi-allelic (homozygous or compound-heterozygous) loss-of-function variation in SNF8. There is no known environmental, infectious, or multifactorial contribution.

The pathogenic principle is quantitative loss of the ESCRT-II holocomplex, not merely loss of SNF8 alone. SNF8 is structurally obligatory for complex integrity, so its depletion co-depletes its partners:

"SNF8 as well as the two physically interacting subunits of ESCRT-II, VPS36, and VPS25 were significantly decreased" (PMID:38423010, patient-derived fibroblasts, individual A2)

Measured fold-changes vs. controls in severe-phenotype fibroblasts: SNF8 0.25 (p=7.8×10⁻¹⁵), VPS36 0.39 (p=0.003), VPS25 0.37 (p=0.005).

Genetic risk factors

• Causal variants: see Section 4. Seven distinct variants across 6 families.
• Susceptibility loci / GWAS: None for this disease. ⚠️ Caution: SNF8 appears in GWAS hits for type 2 diabetes, coronary artery disease, psoriasis, fatigue, and insomnia-cardiac pleiotropy (PMIDs 25469308, 41824857, 40055680, 40176577). These reflect the gene-dense 17q21.32 locus (UBE2Z-GIP-ATP5G1-SNF8) and linkage disequilibrium, not relevance to the Mendelian disorder. Do not curate these as risk factors.
• Modifier genes: None identified. See Section 4 for discussion of why a modifier is hypothesized but unproven.
• Consanguinity: The one homozygous family (Family C, Israeli/Jerusalem, c.623G>T homozygous) is consistent with — but not explicitly reported as — consanguinity. The paper describes families as unrelated to each other; within-family consanguinity is not explicitly stated. Do not assert it.

Environmental risk factors

None known or plausible. No toxin, exposure, lifestyle, occupational, or age/sex risk factor has been reported or is mechanistically implicated. Parental age effects are not applicable (recessive inheritance, not de novo).

Protective factors

None identified. No protective variant, modifier allele, dietary, or lifestyle factor is known.

However, one genuinely important population-genetic observation exists — and it is the single most curation-relevant nuance in this section:

"one apparently healthy individual from the gnomAD population" carries the c.304G>A (p.Val102Ile) variant homozygously (PMID:38423010)

This means the hypomorphic p.Val102Ile allele in the homozygous state is not fully penetrant, implying an unidentified modifier, threshold effect, or ascertainment/phenotyping gap. This is a KNOWLEDGE_GAP discussion candidate, and arguably the most important open question for the entry.

Gene–environment interactions

None reported. No GxE data exist. CTD/PheGenI contain no SNF8 GxE records for this phenotype.

3. Phenotypes

Cohort structure (essential context for all frequencies below)

N=9 total individuals, 6 families. Frequencies are near-meaningless as percentages at this N — I report exact counts and give FrequencyEnum bands only where the count clearly supports one. Per docs/frequency-evidence-guidelines.md, I recommend omitting frequency: on most phenotypes and recording counts in description/notes instead.

Sex ratio: 5 M : 4 F — consistent with autosomal inheritance, no sex bias.

Severe pole (DEE115) — phenotype table

Note on intellectual disability in the severe pole: not formally assessable due to neurodevelopmental arrest and early death. Do not curate HP:0001249 for the severe pole — curate HP:0007281/HP:0001263 instead. This is a real curation trap.

Mild pole (NEDOA + ataxia variant) — phenotype table

Important: ataxia is F1 only — a single individual. The paper's abstract phrase "intellectual disability, childhood-onset optic atrophy, or ataxia" is a disjunction across the mild cohort, not a triad in each patient. Do not curate ataxia as a general feature of NEDOA. F1 also had no optic atrophy and a normal anterior optic tract on MRI at age 2.

Neuroimaging phenotype detail

Severe pole: - "Pronounced and progressive white matter atrophy of the cerebrum" - "Hypo- or aplasia of the corpus callosum beginning at a very early age" - Pachygyria (A2) → suggest HP:0001302 Pachygyria ⚠️ verify - "Rapidly progressive enlargement of the lateral ventricles due to cerebral white matter loss" → HP:0002119 Ventriculomegaly ⚠️ verify - Cerebellum "less severely affected"; brainstem "comparatively normal" — a useful discriminating feature vs. pontocerebellar hypoplasia (see §10 differential)

Mild pole (quantitative volumetry — unusually rich for such a small cohort): - "Severe volume reduction of the intracranial anterior optic pathway including optic nerves (ON), optic chiasma (OC), and optic tracts (2 SD below normal values)" - D1: cerebral white matter −17% bilateral; cerebellar cortex −13% L / −21% R; cerebellar mean diffusivity +13% bilateral - E1: white matter −20% bilateral; temporal cortex −17%; cerebellar cortex −19% L / −17% R; putamen −18% L / −17% R; cerebellar MD +9% L / +8% R - Slight parieto-occipital white-matter hyperintensity - F1 (age 2): normal myelination and white matter volume, slightly dysmorphic callosal body, slight cerebellar atrophy

The increased cerebellar mean diffusivity alongside cortical volume loss is a microstructural signature worth capturing — it supports a degenerative rather than purely dysplastic reading of the mild pole.

Quality-of-life impact

No formal QoL instrument (EQ-5D, SF-36, PROMIS, PedsQL) was administered. No published QoL data exist. Qualitative inference only:

• Severe pole: profound impact — no developmental milestones achieved, gastrostomy dependence, palliative trajectory, death in infancy in 3/4. Total care dependency.
• Mild pole: moderate impact — progressive low vision from childhood (educational and occupational consequence), mild ID predominantly affecting speech/language, independent survival into adulthood (to at least 27 y).

Curate as notes, not as evidenced QoL claims.

4. Genetic / Molecular Information

Causal gene

Pathogenic variants — complete published set (7 variants)

All ✅ confirmed in ClinVar under condition "SNF8-associated disease."

The p.Val102Ile hypomorphic allele — the genotype–phenotype linchpin

This is the single most important genetic fact in the entry:

"All mildly affected individuals shared the same hypomorphic variant, c.304G>A (p.Val102Ile)." (PMID:38423010)

Every one of the four mildly affected individuals (D1, E1, E2, F1) carries p.Val102Ile in trans to a more damaging allele. The severe pole comprises genotypes combining two non-p.Val102Ile alleles. This is a clean, near-deterministic allelic dosage model: residual ESCRT-II function determines pole.

Population and annotation data:

ClinVar submitter disagreement (curate this honestly): - Institute of Human Genetics Munich (2023-12-07): Pathogenic for SNF8-associated disease - Ambry Genetics (2024-01-30): Uncertain significance — "insufficient or conflicting evidence" - OMIM (2024-04-10): Pathogenic for neurodevelopmental disorder plus optic atrophy

The Ambry VUS call is defensible: AF ~1.5×10⁻⁴ with a healthy homozygote is unusual for a fully pathogenic allele. Recommend curating p.Val102Ile with an explicit note on the conflicting classification and reduced penetrance — not as unqualified "Pathogenic."

In silico support (from PMID:38423010)

• CADD for the pathogenic missense variants: 27.3 – 33.0
• REVEL: 0.683 – 0.953
• All variants "affected residues that were spotted in regions documented to be intolerant to variation"
• p.Val102Ile showed "less disruptive impact" by in silico prediction — concordant with its hypomorphic behavior

Functional consequences — mechanism class

Loss of function, acting through two mechanistically distinct routes — an important subtlety:

1. Protein-destabilizing LoF (severe alleles): truncating/frameshift/splice and severe missense alleles reduce SNF8 protein, which co-destabilizes VPS36 and VPS25, collapsing the ESCRT-II holocomplex. Confirmed by quantitative proteomics (fold changes above).

2. Stability-independent LoF (p.Val102Ile): In the mild-phenotype fibroblasts (D1, E1), SNF8 reduction was not statistically significant (D1: 0.74; E1: 0.68) and VPS36/VPS25 were not significantly reduced. The authors conclude the variant:

   "acts via a distinct mechanism independent of protein stability"

No gain-of-function or dominant-negative mechanism is reported. Heterozygous carriers (parents) are unaffected.

gnomAD constraint metrics

⚠️ Not retrieved. The gnomAD GraphQL API was not reachable from this environment (sandbox network restriction) and the gnomAD gene page is a client-rendered JS app that WebFetch cannot resolve. pLI, o/e LoF, mis_z, and LoF observed/expected for SNF8 are therefore NOT reported here — do not populate these fields from memory. Retrieve them directly from https://gnomad.broadinstitute.org/gene/ENSG00000159210 before curating any constraint claim.

Modifier genes

None identified. A modifier is implied by the healthy gnomAD p.Val102Ile homozygote and by the intra-mild-cohort divergence (F1: ataxia, no optic atrophy; D1/E1/E2: optic atrophy, no ataxia — despite all four sharing p.Val102Ile). Curate as a KNOWLEDGE_GAP discussion, not as a finding.

Epigenetic information

None. No methylation episignature, chromatin, or histone-modification data exist for SNF8-related disease. ⚠️ Note a potential confusion source: CHMP1A (a different ESCRT-III gene) has documented chromatin-regulatory function (PMID:23023333) — this does not transfer to SNF8.

Chromosomal abnormalities

None causal. Three ClinVar records retrieved in the SNF8 gene query (VCV003242218, VCV003242217, VCV002692407) are large multi-gene deletions spanning the 17q21.32 region — they are not SNF8-specific and should not be curated as causing this disorder. No recurrent CNV, translocation, or inversion mechanism is known.

5. Environmental Information

Not applicable — no environmental contribution is known or hypothesized.

• Environmental factors: None. CTD contains no SNF8–chemical–disease association for this phenotype.
• Lifestyle factors: None.
• Infectious agents: None causal. ⚠️ Important distinction for curators: the recurrent respiratory infections that caused death in A2 and C2 are a downstream complication of dysphagia and aspiration (i.e., a consequence of the phenotype), not an etiologic or triggering agent. Curate them as complications in Section 11, never as an infectious etiology.

6. Mechanism / Pathophysiology

Causal chain (proposed pathograph, upstream → downstream)

[MOLECULAR] Bi-allelic SNF8 LoF variants
      ↓
[MOLECULAR] Reduced SNF8 protein → co-destabilization of VPS36 + VPS25
    = loss of the ESCRT-II holocomplex
      ↓
[MOLECULAR/CELLULAR] Impaired ESCRT-II → ESCRT-III handoff;
    defective MVB biogenesis + cargo-selective endosomal sorting
      ↓
[CELLULAR] Mis-sorting of lysosomal hydrolases to the (auto)lysosome
      ↓
[CELLULAR] Accumulation of autolysosomes + morphologically aberrant lysosomes
    (enlarged, electron-lucent lumen)
      ↓
[CELLULAR] IMPAIRED AUTOPHAGIC FLUX  ← the convergent hub node
      ↓
[CELLULAR] LC3 accumulation in cortical pyramidal neurons and reactive astrocytes
      ↓
[TISSUE] Myelin loss, reactive gliosis, microglial activation;
 retinal ganglion cell / optic nerve degeneration
      ↓
[TISSUE] Leukoencephalopathy + cerebral atrophy; optic pathway volume loss
      ↓
[ORGANISM] Severe pole: DEE, neurodevelopmental arrest, early death
   Mild pole:   mild ID + childhood-onset optic atrophy / ataxia

Author's own summary of the terminal mechanism (quotable):

"Taken together, we conclude that loss of ESCRT-II due to bi-allelic SNF8 variants is associated with a spectrum of neurodevelopmental/neurodegenerative phenotypes mediated likely via impairment of the autophagic flux." (PMID:38423010)

Note the hedge — "likely via". Autophagic-flux impairment is the authors' preferred hypothesis, well-supported in fibroblasts and neuropathology but not proven to be the proximate cause of the neurological phenotype in vivo. Curate the terminal edge with status: EMERGING in a mechanistic_hypotheses block rather than as established fact.

Molecular pathways

ESCRT / MVB pathway (the core). ESCRT-II is the bridging complex between ubiquitin-cargo recognition (ESCRT-0/I) and membrane scission (ESCRT-III/VPS4).

Structural basis (yeast core, PMID:15329733, Nature 2004 — quotable):

"Here we report the crystal structure of the core of the yeast ESCRT-II complex, which contains one molecule of the Vps protein Vps22, the carboxy-terminal domain of Vps36 and two molecules of Vps25, and has the shape of a capital letter 'Y'. The amino-terminal coiled coil of Vps22 and the flexible linker leading to the ubiquitin-binding NZF domain of Vps36 both protrude from the tip of one branch of the 'Y'."

Note: Vps22 = SNF8. The 1:1:2 stoichiometry (Vps22 : Vps36 : Vps25₂) explains mechanistically why losing SNF8 collapses the whole complex.

ESCRT-II→III coupling (PMID:15469844, Dev Cell 2004 — quotable):

"We show that purified ESCRT-II binds directly to the Vps20 component of ESCRT-III. Surprisingly, this binding does not require the protruding N-terminal coiled-coil of Vps22. Vps25 is the chief subunit responsible for Vps20 recruitment. This interaction dramatically increases binding of both components to lipid vesicles in vitro."

Human ESCRT-II structures: PDB 2ZME, PDB 3CUQ (2.6 Å and 2.9 Å) — three lobes, one copy each of VPS22/VPS36, two copies of VPS25.

Pathways NOT implicated: no evidence for Wnt, MAPK, mTOR, or PI3K-AKT involvement in this disease. Do not import these from generic NDD priors.

Cellular processes

• Macroautophagy / autophagic flux (the central process)
• Multivesicular body biogenesis
• Endosomal cargo sorting; late endosome → lysosome transport
• Lysosomal biogenesis / hydrolase delivery
• Neural progenitor proliferation (zebrafish: ectopic pH3+ cells in forebrain)
• Axon outgrowth/pathfinding (optic nerve extension, optic chiasm formation)
• Reactive gliosis and microglial activation

Cargo selectivity — a mechanistically important negative result

"EGFR degradation was not detectably impaired in patient-derived fibroblasts as compared to fibroblasts from control individuals." (PMID:38423010)

This is significant: EGFR downregulation is the canonical ESCRT-dependent readout, yet it was preserved. The authors attribute this to residual ESCRT-II function and cargo specificity — different cargoes have different ESCRT-II dependency thresholds. Curate this as supports: PARTIAL or as a REFUTE-flavored evidence item against a "global ESCRT collapse" model. It argues for a selective, threshold-dependent defect, which in turn helps explain the mild pole.

Protein dysfunction

SNF8/Q96H20: 258 aa; coiled-coil at residues 27–53; multiple winged-helix repeats (the entire ESCRT-II core is built from eight winged-helix domains, PMID:15469844). Two isoforms via alternative splicing. Interacts with VPS25, VPS36, TSG101, RILPL1, and 14-3-3 proteins.

Mechanism is loss of function via complex destabilization (severe alleles) or stability-independent functional impairment (p.Val102Ile). No misfolding/aggregation proteinopathy is described.

Metabolic changes

None established. ⚠️ Curation trap: PMID:38423010 mentions "mitochondrial complex III deficiency, nuclear type 1" (MIM 124000, BCS1L) only as a separate, unrelated individual retrieved from the same in-house diagnostic database. There is no mitochondrial or complex III defect in SNF8-related disease. No lactate, CSF, or respiratory-chain enzyme abnormality was reported. Do not curate any metabolic/mitochondrial finding.

Immune system involvement

No primary immune involvement. Neuroinflammation is secondary: reactive astrogliosis and microglial activation on autopsy. No autoimmunity, no immunodeficiency.

Tissue damage mechanisms

Autophagic-lysosomal dysfunction → neuronal and oligodendroglial/myelin injury → demyelination and white-matter loss with reactive gliosis. Optic pathway: retinal ganglion cell / optic nerve degeneration (a classic autophagy-vulnerable, long-axon, high-metabolic-demand neuronal population — the same vulnerability class as OPA1/mitochondrial optic neuropathies). No ischemic, fibrotic, or oxidative-stress mechanism is documented.

Biochemical abnormalities

No enzyme deficiency, receptor, ion-channel, or transporter defect. The defect is in vesicular trafficking machinery, not catalysis.

Epigenetic changes

None reported.

Molecular profiling

• Proteomics ✅ — quantitative proteomics on patient fibroblasts is the key functional dataset (SNF8/VPS36/VPS25 fold changes above). Modality: IN_VITRO.
• Transcriptomics — none.
• Metabolomics / lipidomics — none.
• Single-cell / spatial transcriptomics — none.
• Multi-omics integration — none.
• Functional genomics (CRISPR/RNAi screens) — ⚠️ Not retrieved. DepMap was inaccessible (bot-verification wall). Do not assert SNF8 essentiality status without checking https://depmap.org/portal/gene/SNF8 directly. Contextual note only: the IMPC mouse homozygous-null preweaning-lethal phenotype (§15) is consistent with a fitness-critical gene, but that is not the same claim as DepMap common-essentiality.

Suggested GO terms (✅ all verified live)

Biological process | GO ID | Label | |---|---| | GO:0036258 | multivesicular body assembly | | GO:0016236 | macroautophagy | | GO:0061919 | process utilizing autophagic mechanism | | GO:0032456 | endocytic recycling | | GO:0043328 | protein transport to vacuole involved in ubiquitin-dependent protein catabolic process via the multivesicular body sorting pathway |

Cellular component | GO ID | Label | |---|---| | GO:0000814 | ESCRT II complex | | GO:0010008 | endosome membrane | | GO:0031902 | late endosome membrane | | GO:0044754 | autolysosome |

Molecular function | GO ID | Label | |---|---| | GO:0042803 | protein homodimerization activity | | GO:0008289 | lipid binding |

7. Anatomical Structures Affected

Organ level

Primary: central nervous system — cerebrum (white matter predominant), corpus callosum, anterior visual pathway, cerebellum (mild/variable). Secondary: respiratory tract (aspiration pneumonia from dysphagia); GI (feeding failure → gastrostomy); cardiac (terminal arrest during status epilepticus in A1 — an agonal event, not primary cardiac disease). Body systems: nervous (primary), visual/special sense (primary), respiratory + digestive (secondary).

Notably spared: brainstem ("comparatively normal"), and no reported hepatic, renal, hematologic, or skeletal involvement. This sparing pattern is diagnostically useful — it distinguishes SNF8 from the multisystem ESCRT disorder CIMDAG (VPS4A), which features cataracts, anemia, and growth impairment.

Suggested UBERON terms (✅ verified — note label subtleties)

UBERON:0000941 is a genuine trap — its canonical label is "cranial nerve II", so term.label: optic nerve would fail just validate-terms. Use preferred_term: optic nerve with term.label: cranial nerve II.

Tissue and cell level

Affected tissue types: nervous tissue (neurons, glia), myelinated white-matter tracts.

Suggested CL terms

Be explicit in the KB that oligodendrocyte and retinal ganglion cell involvement is inferred, while pyramidal neuron, astrocyte, microglia (autopsy) and fibroblast (in vitro) are directly evidenced.

Subcellular level

Localization / lateralization

Bilateral and symmetric throughout. Volumetric measures were reported bilaterally (e.g., D1 white matter −17% bilaterally; E1 −20% bilaterally), with only minor L/R asymmetry in cerebellar cortex (D1: −13% L vs −21% R). Optic pathway involvement is bilateral (Family C: "bilateral hypoplastic optic nerves"). No unilateral or asymmetric presentation reported. Suggest HP:0012832 Bilateral ⚠️ verify if a laterality modifier is desired.

8. Temporal Development

Onset

Note the bimodal onset distribution — congenital vs. mid-childhood — tracking directly with genotype (p.Val102Ile presence). Onset age is itself a genotype-driven variable here.

Progression

Severe pole — progressive and rapidly fatal: - Stage 1 (birth–~6 mo): profound hypotonia, feeding failure, no milestone acquisition (arrest, not regression) - Stage 2 (~7–9 mo): seizure onset; EEG "multifocal and generalized epileptic discharges further progressing to hypsarrythmia" - Stage 3: emergence of spasticity/hyperreflexia superseding hypotonia (A1, B1) - Stage 4: "pronounced and progressive white matter atrophy," "rapidly progressive enlargement of the lateral ventricles" - Terminal: death 9 wk – 8 mo in 3/4

Rate: rapid. Course: progressive. Duration: fatal in infancy for most; B1 survived to 4.5 y (upper bound of observation, not necessarily of survival).

Mild pole — slowly progressive neurodegeneration on a static developmental baseline: - Static mild ID (developmental, non-progressive) - Superimposed progressive optic neuropathy from age 4–7 y — degenerative, confirmed by abnormal VEPs (E1, E2: "delayed latency and reduced amplitude") and OCT - Slowly progressive volume loss (white matter, cerebellar, temporal cortex, putamen) with increased cerebellar mean diffusivity - Survival to at least 27 y (E1) with no reported deterioration to dependency

Rate: slow. Course: static ID + progressive visual/cerebellar degeneration — a genuine dual-component course. Duration: chronic lifelong.

Terminology precision: in the severe pole this is neurodevelopmental arrest ("Developmental stagnation," HP:0007281) — milestones were never acquired. Do not curate HP:0002376 Developmental regression; the paper reports arrest, not loss of acquired skills.

Patterns

• Remission: none, spontaneous or treatment-induced. No treatment exists.
• Relapsing-remitting: not applicable.
• Critical periods: Mechanistically, the prenatal/perinatal window governs the structural malformation component (corpus callosum hypo-/aplasia, pachygyria, optic nerve hypoplasia) — this is already established at birth and is not modifiable postnatally. The childhood window (4–7 y) is when the progressive optic neuropathy declares itself and is therefore the theoretical window for a disease-modifying (e.g., autophagy-directed) intervention. This is mechanistic reasoning, not published clinical guidance — curate as notes/hypothesis, not as evidence.

9. Inheritance and Population

Epidemiology

Recommended dismech Prevalence record (per the structured-prevalence guidance in CLAUDE.md):

prevalence:
- population: Worldwide
  measure_type: CASES_IN_LITERATURE
  prevalence_class: NOT_YET_DOCUMENTED
  notes: >-
    Nine individuals from six families reported in the single defining
    publication (Brugger et al. 2024). No population prevalence or incidence
    estimate has been published; no Orphanet prevalence class assigned.
  evidence:
  - reference: PMID:38423010
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: "We report nine individuals from six families presenting with a spectrum of neurodevelopmental/neurodegenerative features caused by bi-allelic variants in SNF8"
    explanation: Establishes the total reported case count.

Do not populate rate_per_100000 — there is no basis for a number.

Inheritance

• Pattern: Autosomal recessive — HP:0000007 Autosomal recessive inheritance ⚠️ verify. Confirmed by homozygosity in Family C, compound heterozygosity in 5 families, unaffected heterozygous parents, and unanimous PanelApp classification as "BIALLELIC, autosomal or pseudoautosomal" across all 5 panels.
• Penetrance: Incomplete for at least one genotype. The healthy gnomAD p.Val102Ile homozygote is direct evidence that p.Val102Ile homozygosity is not fully penetrant. For the severe biallelic-null genotypes, penetrance appears complete (4/4 affected + 1 fetal). Curate as incomplete/genotype-dependent, not "complete."
• Expressivity: Highly variable, but largely genotype-explained. The severe↔mild split maps almost deterministically onto p.Val102Ile presence. Residual intra-mild variability (F1 ataxia-without-optic-atrophy vs. D1/E1/E2 optic-atrophy-without-ataxia) is unexplained.
• Genetic anticipation: Not applicable — not a repeat-expansion disorder.
• Germline mosaicism: Not reported.
• Founder effects: None established. Family C's homozygous c.623G>T (Israeli/Jerusalem) is a single family — insufficient to claim a founder allele. Do not assert one. The recurrence of p.Val102Ile across three unrelated families (D, E, F — two Italian, one German) reflects its appreciable population frequency (a segregating polymorphic hypomorph), not a founder event.
• Consanguinity: Not explicitly reported; plausible in Family C given homozygosity. Do not assert.

Population demographics

• Reported ancestry: German (Families A/Munich, B/Leipzig, F/mixed European), Italian (D/Bologna, E/Rome), Israeli (C/Jerusalem). Entirely European + Middle Eastern — this reflects ascertainment through European diagnostic-genomics networks, not true population restriction. Explicitly flag ascertainment bias; do not curate "affects Europeans."
• Geographic distribution: No endemic pattern; ascertainment-driven.
• Variant geography: p.Val102Ile in Italian and German families; c.623G>T homozygous in the Israeli family. N is far too small for geographic inference.
• Sex ratio: 5 M : 4 F — no sex bias, as expected for autosomal recessive.
• Age distribution: bimodal by pole — severe pole ascertained in infancy (9 wk – 8 mo at death; B1 to 4.5 y); mild pole ascertained in childhood through adulthood (4–27 y).

10. Diagnostics

Clinical tests

Laboratory tests / biomarkers: No diagnostic biochemical biomarker exists. No lactate, CSF, enzyme-assay, or metabolic marker abnormality was reported. Diagnosis is molecular-genetic, full stop. ⚠️ Do not curate a metabolic screen as diagnostic.

Research-only functional assays (not clinically validated, IN_VITRO): - Quantitative proteomics on patient fibroblasts showing reduced SNF8/VPS36/VPS25 — the most useful functional confirmation for a VUS - Electron microscopy: enlarged vesicular structures containing cytoplasmic material; lysosomes with "enlarged size and a largely electron lucent lumen" - LC3/LAMP1 immunofluorescence: autolysosome accumulation

Imaging — brain MRI (the key diagnostic modality; suggest NCIT:C16809 Magnetic Resonance Imaging ⚠️ verify): - Severe: progressive cerebral white-matter atrophy, corpus callosum hypo-/aplasia, ventriculomegaly, pachygyria; cerebellum less affected, brainstem comparatively normal - Mild: quantitative volumetry of the anterior visual pathway (ON/OC/optic tracts ≥2 SD below normal) — the highest-yield mild-pole finding; plus regional volume loss and increased cerebellar mean diffusivity (DTI)

Functional / electrophysiology: - EEG: multifocal and generalized epileptic discharges "further progressing to hypsarrythmia" (severe pole) - Visual evoked potentials (VEP): "delayed latency and reduced amplitude" (E1, E2) — sensitive for the mild pole - OCT: confirms optic (hypo)atrophy, onset 4–7 y

Biopsy / pathology: Not diagnostic. Autopsy neuropathology (A2) is confirmatory/research: pachygyria, enlarged sulci, reduced cerebral white matter, corpus callosum thinned to 2 mm, "marked loss of myelin," reactive astrogliosis, microglial activation, and strong LC3 immunostaining in internal pyramidal cell layer neurons and reactive white-matter astrocytes (vs. "only weak staining" in age-matched controls).

Genetic testing — the definitive route

Recommended approach: WES or WGS (or a broad NDD/epilepsy/optic-neuropathy panel including SNF8), followed by phase/segregation confirmation in parents to establish biallelic status. SNF8 is a small gene (258 aa) and single-gene testing is not clinically offered.

Genomics England PanelApp — SNF8 appears on 5 panels (all "BIALLELIC, autosomal or pseudoautosomal"):

The Optic neuropathy amber rationale: "two unrelated cases reported with optic atrophy" — i.e., below the green threshold for that specific phenotype. This Green/Amber split is genuinely informative and worth capturing: SNF8 is diagnostic-grade for ID/DD and fetal anomalies, but only moderate-evidence for isolated optic neuropathy and epilepsy presentations.

Other modalities: - CMA: not indicated (no CNV mechanism) - Karyotype / FISH: not indicated - mtDNA testing: not indicated — but see differential below; it is often performed before the diagnosis because optic atrophy prompts mitochondrial workup - Repeat expansion testing: not applicable

Omics-based diagnostics

• RNA-seq: ⚠️ Genuinely worth flagging as a curation-relevant gap. The c.423−1G>C splice-acceptor variant (Family D) is annotated p.? — its actual transcript consequence was not experimentally determined. RNA-seq/RT-PCR would resolve it. This is a concrete KNOWLEDGE_GAP candidate.
• Proteomics: research-grade functional support (see above); not a clinical test.
• Metabolomics / epigenomics (episignature) / liquid biopsy: none available.

Clinical criteria

No consensus diagnostic criteria exist (single publication, N=9). Practical diagnosis = compatible phenotype + biallelic SNF8 variants + parental segregation.

Differential diagnosis

For the severe pole (DEE + leukoencephalopathy + CC hypo-/aplasia):

For the mild pole (ID + childhood-onset optic atrophy):

Key practical point: the mild pole most closely mimics ACO2/Behr-type optic atrophy + ataxia + mild ID syndromes, and mitochondrial optic neuropathies are typically excluded first. SNF8 is likely under-ascertained in exome-negative optic-atrophy-plus-ID cohorts — a plausible reason the published N remains 9.

Screening

• Newborn screening: Not included in any program; no biochemical marker exists, so NBS is not technically feasible by current MS/MS methods.
• Carrier screening: SNF8 is not on standard expanded carrier panels. Given AF ~1.5×10⁻⁴ for p.Val102Ile alone and unresolved penetrance, inclusion is not currently justified.
• Cascade screening: Appropriate within families — targeted testing of at-risk relatives and reproductive partners once a familial genotype is known.

11. Outcome / Prognosis

Survival and mortality

⚠️ No survival curve, 5-/10-year survival rate, life expectancy estimate, or mortality rate has been published. With N=9, only individual outcomes are reportable. Do not compute or curate a percentage survival figure — any such number would be fabricated.

Observed outcomes (individual-level, PMID:38423010):

Disease-specific mortality mechanisms — both are secondary complications, which is prognostically actionable: 1. Aspiration/respiratory infection from dysphagia (2 of 3 deaths) — the leading cause 2. Status epilepticus (1 of 3 deaths)

Morbidity and function

• Severe pole: total functional dependency; no developmental milestones achieved; gastrostomy dependence (3/4); progressive spastic tetraplegia. Profound disability.
• Mild pole: mild ID predominantly affecting speech/language; progressive low vision from childhood; F1 with congenital ataxia. Substantially preserved function — all mild-pole individuals survived to adolescence/adulthood.
• ICF-coded disability outcomes / DALYs: none published.
• Quality of life instruments: none administered (see §3).

Complications

Recovery potential

None. No recovery, no reversal of established structural CNS damage. The malformation component (CC hypo-/aplasia, pachygyria, optic nerve hypoplasia) is prenatally established and irreversible. No treatment exists to alter the degenerative component.

Prognostic factors

The dominant prognostic factor is genotype — specifically, presence of p.Val102Ile:

"All mildly affected individuals shared the same hypomorphic variant, c.304G>A (p.Val102Ile)." (PMID:38423010)

This is a genuinely strong, near-deterministic genotype–prognosis correlation in the published cohort — presence of one p.Val102Ile allele in trans predicted the mild pole in 4/4 cases; its absence predicted the severe pole in 5/5 (including the fetus). Caveat clearly: N=9, single cohort, no independent replication.

Supporting functional/prognostic correlates: - Degree of ESCRT-II subunit depletion in fibroblasts tracks severity (severe: SNF8 0.25 with significant VPS36/VPS25 loss; mild: SNF8 0.68–0.74, non-significant, VPS36/VPS25 preserved) — a candidate functional prognostic assay - Zebrafish allele-pair severity recapitulated the human gradient: severe pair (p.Tyr167Ter + p.Gly191Asp) ~95% aberrant embryos vs. mild pair (p.Pro79Leu + p.Val102Ile) ~70% - Presence of dysphagia/gastrostomy dependence and early seizure onset mark the poor-prognosis group clinically

Prognostic biomarkers: none clinically validated.

12. Treatment

⚠️ There is NO disease-specific or disease-modifying treatment for SNF8-related neurodevelopmental disorder.

PMID:38423010 discusses no therapeutic intervention. Verbatim finding from full-text review: "No therapeutic interventions, treatments, or management strategies are discussed in this paper." Severely affected individuals received palliative care; three required gastric tube feeding.

A search of ClinicalTrials.gov-indexed literature and the publication record identified NO clinical trials, NO NCT identifiers, and NO experimental therapeutics for this disorder. Any treatment content below is standard-of-care symptomatic management inferred from the reported clinical needs, and must be curated as such — with notes rather than fabricated evidence, or with evidence citing only what PMID:38423010 actually states (i.e., that gastrostomy feeding was required).

Pharmacotherapy

• Antiseizure medications for the epileptic encephalopathy. No specific agent, regimen, or response rate is reported; the epileptic encephalopathy with hypsarrhythmia was clinically severe and A1 died in status epilepticus, suggesting poor pharmacoresponsiveness — but the paper does not state this. Do not invent drug names.
• Suggested: treatment_term NCIT:C15986 Pharmacotherapy; therapeutic_modality: SMALL_MOLECULE. Leave therapeutic_agent empty — no agent is documented.
• Pharmacogenomics: none. No PharmGKB/CPIC entry relates to SNF8.

Advanced therapeutics

None exist. No gene therapy, gene editing, cell therapy, ASO/siRNA/mRNA therapy, targeted therapy, or immunotherapy has been developed, trialed, or proposed in print.

Theoretical considerations only — flag clearly as speculative if curated at all: recessive LoF with a hypomorphic-allele-defines-mild-pole architecture is in principle gene-replacement-tractable, and the p.Val102Ile natural experiment suggests only partial restoration of ESCRT-II function may suffice for the mild phenotype — a favorable therapeutic-threshold argument. But: the prenatally established malformation component (CC hypo-/aplasia) sets a hard limit on postnatal benefit, and CNS-wide delivery to white matter and retinal ganglion cells is unsolved. This is my mechanistic reasoning, not published work — do not curate as evidence.

Surgical and interventional

• Gastrostomy / PEG placement — the one intervention actually documented: 3 of 4 severely affected individuals required gastric tube feeding.
• Suggested: NCIT:C15329 Surgical Procedure (or a specific gastrostomy term, ⚠️ requires OAK verification); therapeutic_modality: SURGERY
• Quotable basis: the paper states three required "gastric tube feeding"

Supportive and rehabilitative (the actual standard of care)

Per the CLAUDE.md mechanical-backfill table: NCIT:C15302 → BEHAVIORAL, NCIT:C15329 → SURGERY, NCIT:C15986 → agent-dependent (do not auto-assign). NCIT:C15433 Nutritional Support must NOT be mechanically tagged BEHAVIORAL — this exact mis-tagging was tried and reverted in this repo on 2026-07-08.

Treatment outcomes, algorithms, combination/personalized approaches

None published. No response rates, no adverse-event data (no disease-specific drug exists), no treatment algorithm, no NCCN/society guideline, no genotype-guided treatment protocol. The only genotype-driven clinical action is prognostic counseling (p.Val102Ile → mild pole expectation), not treatment selection.

13. Prevention

Primary prevention

Not preventable — a germline monogenic disorder. The only primary-prevention modality is reproductive: - Genetic counseling with 25% recurrence risk per pregnancy for carrier couples (NCIT:C15240 Genetic Counseling) - Prenatal diagnosis (CVS/amniocentesis) with targeted testing of the known familial variants - Preimplantation genetic testing for monogenic disorders (PGT-M) — technically applicable once familial variants are known - ⚠️ Counseling caveat: for couples where the fetus would be a p.Val102Ile homozygote, counseling is genuinely uncertain given the healthy gnomAD homozygote and the Ambry VUS classification. This must be communicated as uncertain, not as "affected." Important nuance to capture.

Family C's history — termination of pregnancy at 25 weeks for detected brain malformations, followed by an affected liveborn sibling (C2) who died at 9 weeks — illustrates the real reproductive stakes.

Secondary prevention (early detection)

• Cascade genetic testing of at-risk relatives in known families
• In the mild pole, serial ophthalmologic surveillance (VEP, OCT, visual acuity) from early childhood is a rational early-detection strategy given documented onset at 4–7 y — enabling timely low-vision support and educational accommodation. This is inference from the reported onset window, not a published guideline.

Tertiary prevention (complication prevention) — the highest-yield real intervention

Given that 2 of 3 deaths were respiratory infections secondary to dysphagia, tertiary prevention is where meaningful clinical benefit plausibly lies: - Early swallow assessment and aspiration-risk management - Timely enteral feeding (gastrostomy) before nutritional/respiratory decompensation - Respiratory hygiene, immunization against respiratory pathogens (routine schedule), prompt infection treatment - Seizure-emergency (status epilepticus) action plans — relevant given A1's death - Contracture prevention via physiotherapy/positioning

Immunization

No disease-specific vaccine. Routine childhood immunization — with attention to respiratory pathogens (influenza, pneumococcus, RSV) — is rational given the aspiration-pneumonia mortality pattern. Standard-of-care inference, not published guidance.

Screening programs, risk stratification, behavioral interventions, public health, prophylaxis

• Population screening: not warranted (ultra-rare, no biomarker, no intervention)
• Newborn screening: not feasible (no biochemical marker)
• Risk stratification: genotype-based only (p.Val102Ile presence → mild pole)
• Behavioral / lifestyle interventions: none applicable — no modifiable lifestyle risk factor exists
• Public health / environmental interventions: not applicable
• Prophylaxis: no antimicrobial or other prophylaxis protocol published; aspiration-pneumonia prevention is the rational target

14. Other Species / Natural Disease

Taxonomy and orthologs

⚠️ I attempted to confirm the mouse MGI accession and hit a wrong record (MGI:1913677 is Cyb5b, not Snf8). Do not curate MGI:1913677. The MGI ID for Snf8 must be looked up fresh before use. ZFIN quick-search returned 404 in this environment; the ZFIN ID likewise needs direct verification.

Breed (VBO)

Not applicable — no breed-associated natural disease.

Natural disease in other species

None known. No naturally occurring SNF8-related disease has been reported in companion animals, livestock, or wildlife. A targeted OMIA search returned no SNF8 entry. No veterinary relevance.

⚠️ Do not curate the following as animal disease models — they are incidental GWAS/biomarker associations in the SNF8-containing locus and are unrelated to this disorder: - Chicken carcass-weight GWAS (PMID:40211845) - Drosophila sleep/cardiac pleiotropy knockdown (PMID:40176577) - C. elegans longevity/DAF-16 (PMID:32829877) - Fathead minnow viral hemorrhagic septicemia proteomics (PMID:24931624) - Rat MASH models (PMID:40306176)

Comparative biology

Evolutionary conservation is strong and mechanistically meaningful. ESCRT-II is conserved from yeast to human: "ESCRT-II plays a pivotal role in receptor downregulation and multivesicular body biogenesis and is conserved from yeast to humans." The subunit architecture is preserved — yeast Vps22/Vps36/Vps25 ↔ human SNF8/VPS36/VPS25, with the same 1:1:2 stoichiometry and Y-shaped/trilobal fold (PMID:15329733; PMID:15469844; human structures PDB 2ZME, 3CUQ).

Comparative pathology: the human disease phenotype (CNS-restricted neurodevelopmental/neurodegenerative) is not recapitulated as a natural disease in any species. Yeast and invertebrate models display trafficking, gene-expression, and longevity phenotypes with no neurodevelopmental correlate — the yeast literature (glucose-dependent gene expression, Rim101/PMR1 calcium-pump regulation, PAF1-complex genetic interactions, flavor-ester biosynthesis) reflects conserved machinery in a non-conserved physiological context. Useful for structure–function, not for disease modeling.

Transmission

Not applicable — no zoonotic potential, no cross-species susceptibility, not communicable.

15. Model Organisms

Zebrafish (Danio rerio) — the primary in vivo disease model

Model type: vertebrate, morpholino (MO) antisense knockdown — ⚠️ transient knockdown, NOT a stable genetic mutant. This distinction matters for evidence weighting.

Recapitulated phenotypes (from the abstract, quotable):

"Snf8 loss of function in zebrafish results in global developmental delay and altered embryo morphology, impaired optic nerve development, and reduced forebrain size."

Detailed findings: - Statistically significant global developmental delay, curly tail, reduced pigmentation, small head and eyes - Reduced brain area on confocal morphometry, "with variable phenotypes among morphant fish" - Ectopic proliferative (pH3+) cells within the forebrain, "partially rescued in fish expressing WT SNF8" → implicates dysregulated neural progenitor proliferation - Optic nerve: statistically significant reduction in extension and thickness, "which was rescued in fish co-injected with snf8 MO and WT SNF8"; reduced axonal scaffold - Optic chiasm: loss of the characteristic angle in morphants, restored by WT SNF8

Variant-specific rescue — the key pathogenicity and genotype–phenotype experiment: - WT SNF8 mRNA partially rescued the phenotype - Disease-variant allele pairs failed to rescue - Severity gradient mirrored the human poles:

"A more severe phenotypic impact in embryos coexpressing the SNF8 alleles encoding p.Tyr167Ter and p.Gly191Asp compared to those microinjected with alleles encoding p.Pro79Leu and p.Val102Ile was observed (approximately 95% vs. 70% aberrant embryos)."

Phenotype recapitulation quality — genuinely good for the axis that matters: the model reproduces (a) impaired optic nerve/chiasm development ↔ human optic atrophy/optic nerve hypoplasia, and (b) reduced forebrain size ↔ human microcephaly/reduced cerebral volume. Critically, it reproduces the allele-severity gradient, providing independent in vivo support for the p.Val102Ile-hypomorph model.

Limitations (curate honestly): 1. Morpholino, not a germline mutant — subject to well-known off-target/toxicity artifacts; no stable snf8 zebrafish mutant line is reported. Rescue experiments mitigate but do not eliminate this concern. 2. Does not model epilepsy — no seizure phenotype assessed; the DEE component is unmodeled 3. Does not model leukoencephalopathy or corpus callosum hypo-/aplasia — zebrafish lack a corpus callosum entirely 4. Does not model intellectual disability or the progressive postnatal neurodegenerative course 5. Variable penetrance among morphants ("reduced penetrance of morphological defects")

Recommended dismech treatment: curate zebrafish evidence with evidence_source: MODEL_ORGANISM, and add a kind: HUMAN_MODEL_MISMATCH discussion — not a generic KNOWLEDGE_GAP — since evidence exists in the model but its fidelity to the human phenotype is the open question (per the CLAUDE.md distinction: KNOWLEDGE_GAP = evidence absent; HUMAN_MODEL_MISMATCH = evidence exists but translational validity is uncertain). The specific mismatch: the zebrafish MO model captures the optic-nerve/forebrain developmental axis but captures none of the epileptic encephalopathy, callosal agenesis, or leukoencephalopathy that define the severe human pole.

Mouse (Mus musculus) — IMPC knockout

Retrieved live from the IMPC genotype-phenotype API for Snf8:

Interpretation: Snf8 homozygous null is preweaning lethal with complete penetrance in mouse — consistent with an essential gene and concordant with the severity of the human biallelic-null phenotype (death in infancy in 3/4). This is a genuinely useful, independently-sourced cross-species datapoint.

Limitations: 1. Complete null lethality precludes study of the postnatal neurological phenotype — no viable homozygous mouse exists to phenotype for seizures, white matter, or optic nerve 2. No hypomorphic mouse allele exists — the human p.Val102Ile mild pole, which is the most clinically informative genotype, is entirely unmodeled in mouse. A knock-in Snf8 p.Val102Ile equivalent is the obvious highest-value missing model. Strong proposed_experiments candidate. 3. Heterozygous phenotypes (tail elevation, elevated LDH) have no clear human correlate — human heterozygous carriers (parents) are unaffected. Do not over-interpret. 4. No conditional, neural-specific, or humanized mouse model is reported.

⚠️ Verify the MGI accession for Snf8 directly before curating (my first lookup landed on Cyb5b).

Cellular / in vitro models

Patient-derived primary fibroblasts — the workhorse functional system (individuals A2 [severe], D1 and E1 [mild]); evidence_source: IN_VITRO: - Quantitative proteomics (ESCRT-II subunit quantification) - Transmission electron microscopy (autolysosome and aberrant lysosome accumulation) - LC3/LAMP1 confocal immunofluorescence - EGFR degradation assay (preserved — the cargo-specificity result)

Limitation: fibroblasts are not neurons. The disease is CNS-restricted, so the most disease-relevant cell types — cortical pyramidal neurons, oligodendrocytes, retinal ganglion cells — were not studied functionally. No iPSC, iPSC-derived neuron, organoid, or immortalized-line model is reported. This is a significant gap: iPSC-derived neurons and retinal organoids are the obvious next models, and were used in the comparable VPS4A work (PMID:33186545 studied "iPSC-derived human neurons"), showing the approach is tractable for ESCRT disorders.

MorPhiC: SNF8 is not among the MorPhiC anchor genes (ISL1, EOMES, GCM1, NKX2-1). No MorPhiC cellular phenotype data available.

Induced (non-genetic) models

None — no drug-induced, surgical, or environmentally induced model exists or would be meaningful for a monogenic trafficking defect.

Model databases to query

MGI (mouse — verify Snf8 accession), IMPC/mousephenotype.org (✅ data retrieved above), ZFIN (zebrafish — verify accession), FlyBase (Vps22), WormBase (vps-22), SGD (yeast SNF8), Alliance of Genome Resources (cross-species), IMSR/KOMP/EMMA/MMRRC (strain availability — ⚠️ not checked).

Appendix A — Curation notes on the existing draft entry

I found an untracked draft at kb/disorders/SNF8-Related_Neurodevelopmental_Disorder.yaml in this worktree. Five observations, most-actionable first:

1. updated_date should be removed. CLAUDE.md states plainly: "Do not add updated_date to new entries. The field is deprecated — git history is the authoritative change log." The draft sets updated_date: "2026-08-01T00:00:00Z".

2. Scope mismatch between the entry name and disease_term. The entry is named "SNF8-Related Neurodevelopmental Disorder" (spanning both poles) but binds disease_term to MONDO:0968947 (neurodevelopmental disorder plus optic atrophy = the mild pole only), and its description describes only "the milder end." The severe pole — DEE115, MONDO:0968946 / OMIM:620783 — carries 4 of 9 individuals and all the mortality, and is currently unrepresented. Two clean options:

3. (a) Keep one spectrum entry, use has_subtypes with Severe (DEE115) and Mild (NEDOA) as subtype names (short and slug-friendly per the naming convention), and add MONDO:0968946 under mappings:; or
4. (b) Split into two Disease entries and add a Grouping over them (grouping_basis: SHARED_GENE_FAMILY + SHARED_MECHANISM, with a NECESSARY HAS_GENE criterion on SNF8).

I'd recommend (a) — the poles are a genotype-graded continuum of one mechanism, not two diseases, and the has_subtypes foreign-key machinery lets phenotypes/prevalence/progression be attributed per pole.

1. No evidence: blocks anywhere in the draft. Every pathophysiology node, phenotype, and genetic claim needs an EvidenceItem. PMID_38423010.md is already in references_cache/ with the full abstract, so exact-quote snippets can be drawn from it immediately without a fetch. Verbatim snippets ready to use (all confirmed substrings of the cached abstract):
2. "All mildly affected individuals shared the same hypomorphic variant, c.304G>A (p.Val102Ile)."
3. "In patient-derived fibroblasts, bi-allelic SNF8 variants cause loss of ESCRT-II subunits."
4. "Snf8 loss of function in zebrafish results in global developmental delay and altered embryo morphology, impaired optic nerve development, and reduced forebrain size."
5. "loss of ESCRT-II due to bi-allelic SNF8 variants is associated with a spectrum of neurodevelopmental/neurodegenerative phenotypes mediated likely via impairment of the autophagic flux"

6. "The phenotypic spectrum included four individuals with severe developmental and epileptic encephalopathy, massive reduction of white matter, hypo-/aplasia of the corpus callosum, neurodevelopmental arrest, and early death."

7. hgnc:17028 is correct for SNF8 (lowercase prefix, per repo convention) — ✅ verified against the HGNC REST API.

8. The single pathophysiology node should be expanded into the causal chain in §6. The most valuable addition is the convergent hub node "Impaired Autophagic Flux" with biological_scale: CELLULAR, since that is the mechanistic claim the paper actually makes — and it should carry hypothesis_groups reflecting the authors' "likely via" hedge (a mechanistic_hypotheses entry with status: EMERGING).

Module conformance opportunity: no existing kb/modules/ module cleanly fits (ESCRT/autophagic-flux is not yet modeled; lysosomal_substrate_accumulation is a hydrolase-deficiency/substrate-storage module and is not the right anchor here — there is no stored substrate). Candidate future modules: an ESCRT/autophagic-flux failure module, or photoreceptor_degeneration's sibling for optic neuropathy. Worth noting there is a real gap rather than forcing a bad conforms_to.

Appendix B — Recommended discussions entries

Sources

Primary (disease-defining): - Brugger M, et al. Bi-allelic variants in SNF8 cause a disease spectrum ranging from severe developmental and epileptic encephalopathy to syndromic optic atrophy. Am J Hum Genet. 2024;111(3):594–613. PMID:38423010 · PMC full text · ScienceDirect

Mechanism / structural biology: - Teo H, et al. Structure of the ESCRT-II endosomal trafficking complex. Nature. 2004. PMID:15329733 - Hierro A, et al. ESCRT-II, an endosome-associated complex required for protein sorting: crystal structure and interactions with ESCRT-III and membranes. Dev Cell. 2004. PMID:15469844 - Structure and function of the ESCRT-II-III interface in multivesicular body biogenesis. Dev Cell. 2009. PMID:19686684 - RCSB PDB 2ZME — human ESCRT-II complex · PDB 3CUQ - UniProt Q96H20 — Vacuolar-sorting protein SNF8

Related ESCRT disorders (differential diagnosis): - Mochida GH, et al. CHMP1A encodes an essential regulator of BMI1-INK4A in cerebellar development. Nat Genet. 2012. PMID:23023333 - Rodger C, et al. De Novo VPS4A Mutations Cause Multisystem Disease with Abnormal Neurodevelopment. Am J Hum Genet. 2020. PMID:33186545

Structured databases: - OMIM 620783 — DEE115 · OMIM 620784 — NEDOA · OMIM 610904 — SNF8 gene - ClinVar SNF8 variants (VCV002664478–VCV002664484) - HGNC:17028 (REST) - MONDO via EBI OLS4 — MONDO:0968946, MONDO:0968947 - HPO term verification via JAX HPO API - IMPC genotype-phenotype API — Snf8 - Genomics England PanelApp — SNF8 (Optic neuropathy, panel 186) · all SNF8 panel entries - GeneCards — SNF8

Verified unavailable (searched, nothing found): Orphanet ORPHA code · ClinicalTrials.gov trials · OMIA natural animal disease · MorPhiC data · published QoL/survival/prevalence statistics · GxE data · episignature data · any second clinical publication.

Not retrievable in this environment (do NOT populate from memory — fetch directly): gnomAD constraint metrics (pLI, o/e LoF) for SNF8 · DepMap essentiality · MGI accession for mouse Snf8 · ZFIN accession for zebrafish snf8 · OMIM full clinical synopses (omim.org returned HTTP 403).