Epilepsy Granularity and Mechanism Curation Project

Active Artifacts

• Epilepsy project brief
• Mechanism landscape and literature map
• Initial candidate disorder scope
• Temporal lobe epilepsy deep research note
• MTLE-HS deep research note

Overview

This project records a practical curation plan for representing epilepsy in dismech at a level where mechanisms are coherent and clinically useful.

It is based on current dismech granularity discussions and pathophysiology debundling work: - https://github.com/monarch-initiative/dismech/issues/272 - https://github.com/monarch-initiative/dismech/issues/306 - https://github.com/monarch-initiative/dismech/issues/164 - https://github.com/monarch-initiative/dismech/issues/332 - https://github.com/monarch-initiative/dismech/pull/210

Core Granularity Rules

Split into separate entries when: - Pathophysiology is distinct. - Treatment response differs (especially precision/targeted treatment implications). - Prognosis or disease course differs materially. - External disease resources treat entities as distinct.

Lump into one entry (with has_subtypes) when: - Core mechanism is shared. - Phenotypes and management substantially overlap. - Splitting does not add practical curation/clinical value.

Pathophysiology modeling rule: - Keep mechanisms atomic and causally linked. - Do not mix phenotypes into mechanism events.

Current KB Anchors

• Umbrella entry: kb/disorders/Epilepsy.yaml (MONDO:0005027)
• Has broad subtype list at kb/disorders/Epilepsy.yaml:10.
• Contains mechanism blocks that are partially subtype-specific (for example TLE/mTOR evidence), see kb/disorders/Epilepsy.yaml:44, kb/disorders/Epilepsy.yaml:48, kb/disorders/Epilepsy.yaml:136.
• Existing mechanistically coherent subtype-level entries:
• kb/disorders/Dravet_syndrome.yaml (MONDO:0100135) with a clear SCN1A -> interneuron dysfunction -> hyperexcitability chain at kb/disorders/Dravet_syndrome.yaml:17 and kb/disorders/Dravet_syndrome.yaml:50.
• kb/disorders/Jeavons_Syndrome.yaml (MONDO:0015346) with phenotype coherence but mixed-quality mechanism evidence (examples at kb/disorders/Jeavons_Syndrome.yaml:50, kb/disorders/Jeavons_Syndrome.yaml:105, kb/disorders/Jeavons_Syndrome.yaml:149).

Starting Points (Mechanism-Coherent Lumps)

1. Temporal lobe epilepsy (TLE) as its own entry

• Why: current umbrella epilepsy evidence is partly TLE-specific.
• Evidence anchors: kb/disorders/Epilepsy.yaml:44, kb/disorders/Epilepsy.yaml:48, kb/disorders/Epilepsy.yaml:189.
• Proposed initial target file: kb/disorders/Temporal_Lobe_Epilepsy.yaml
• Status: [ ] Planned

2. mTOR-related focal epilepsy

• Why: coherent pathway-level mechanism with therapeutic relevance (mTOR inhibition).
• Evidence anchors: kb/disorders/Epilepsy.yaml:122, kb/disorders/Epilepsy.yaml:136, kb/disorders/Epilepsy.yaml:144.
• Proposed initial target file: kb/disorders/mTOR_related_focal_epilepsy.yaml
• Status: [ ] Planned

3. BBB-disruption-associated drug-resistant focal epilepsy

• Why: coherent endothelial/BBB mechanism with translational signal.
• Evidence anchors: kb/disorders/Epilepsy.yaml:50, kb/disorders/Epilepsy.yaml:66, kb/disorders/Epilepsy.yaml:69.
• Proposed initial target file: kb/disorders/BBB_disruption_associated_focal_epilepsy.yaml
• Status: [ ] Planned

4. SCN1A-related epileptic encephalopathy (Dravet spectrum)

• Why: already coherent and mostly in place; use as reference model for future sodium-channel DEE entries.
• Current anchor: kb/disorders/Dravet_syndrome.yaml
• Status: [ ] Consolidate and template

5. Photosensitive eyelid-myoclonia epilepsy family (Jeavons/EMA)

• Why: coherent electroclinical phenotype family; good candidate after evidence cleanup.
• Current anchor: kb/disorders/Jeavons_Syndrome.yaml
• Status: [ ] Cleanup first, then consider family-level structure

Practical Modeling Decision for dismech

• Keep Epilepsy (MONDO:0005027) as umbrella/index-level entry.
• Move mechanism-specific causal chains into subtype-level entries where coherence is better.
• Keep umbrella entry concise and focused on cross-cutting mechanisms only.

Initial Execution Plan

1. Create Temporal_Lobe_Epilepsy.yaml by refactoring TLE-specific content out of umbrella epilepsy.
2. Create mTOR_related_focal_epilepsy.yaml using the mTOR and dysmorphic neuron chain.
3. Keep Dravet_syndrome.yaml as the template for mechanistically coherent epilepsy subtype modeling.
4. Perform evidence quality pass on Jeavons_Syndrome.yaml before additional splitting.

Notes

2026-09-10

• Ran a scripted census of every epilepsy entry (scripts/pediatric_epilepsy_census.py -> research/pediatric_epilepsy_census.md): 109 entries in the MONDO epilepsy closure, 84 of them pediatric, plus 36 entries the KB treats as epilepsies from outside the closure and 46 with seizures as an obligate or very frequent phenotype.
• Coverage of the AAP pediatric epilepsy types is 20 of 23 complete; reflex epilepsies, self-limited neonatal-infantile seizures, and sleep-related hypermotor epilepsy are partial.
• The gap is mechanism, not coverage: 91 entries conform to epilepsy_excitation_inhibition_imbalance and for 78 that is the only module. Added the Mechanisms of the Epilepsies module collection and proposed five new modules; see docs/reports/pediatric-epilepsy-census-2026-09-10.md.

2026-02-16

• Recorded initial epilepsy granularity strategy and starting candidates.
• Strategy aligns with active granularity issues and prior pathophysiology debundling direction.