WWOX: what to study, and on what evidence

Main investigations to pursue

Order of the investigations to be funded: it assesses the value of the next discriminating test, not the efficacy of an intervention. Provisional and reasoned order, with possible ties: each row is a question about a compound or intervention, its action, the context and the benefit sought, with a next test. It is not a ranking of treatments nor a probability of benefit; the depth of the evidence is indicated separately.

1. 1 (tie) AAV9 gene addition of human WWOX (gene therapy; see the gene-addition cards) thread already known in WWOX thread with a result already in a card#

   Question

   Does neuronal WWOX restoration started AFTER symptom onset recover any endpoint, or only prevent one when given in the neonatal window?

   Reason for the priority

   Every partial-benefit lever in this module is ordered as a bridge to, or a substitute for, restoration; whether restoration works after onset decides which of those two framings is the real one, so one result re-reads the whole list.

   Solidity of the evidence (depth of reading)

   E4 - card reviewed, 12 WWOX-side assertion(s) read beyond abstract; class pharmacology table (IT7 P5)

   Decisive uncertainty

   No arm exists in which treatment begins after the phenotype is established; the window is stated, not measured against a later start, and P300 readouts are survivor-selected.

   Next test

   One added arm of the published design with injection after the first recorded spike-wave discharge, same endpoints, untreated same-genotype comparator kept. Worded per endpoint class: measure the restoration actually achieved at the age of treatment and its regional distribution, because the one comparable later-age failure in a short-lived null was left uninterpretable by that missing measurement; carry a within-animal pre-treatment recording, a treated wild-type or heterozygote arm, growth and glycaemia as primary endpoints (growth failed even with pre-onset treatment in the closest analogue), and follow-up long enough to see decay. Quantified myelin sheath structure (g-ratio or electron microscopy, not marker intensity) is the discriminating endpoint for a neuron-driven lesion: white-matter abnormality has been reversed after a documented onset only in glial or metabolic lesions, and sheath structure after onset has been reported only in a metabolic oligodendrocyte-enzyme lesion (G-ratio by electron microscopy, one source) and not in any neuron-driven lesion. Spike-wave rate is the confirmatory endpoint and its external precedent (not WWOX evidence) is split, so pair it with a rhythm or power-spectrum measure; interneuron number as well as excitability, because no source measured number.

   Scientific details of this row

   Origin of the proposal

   Named in the WWOX literature and carded: the neonatal AAV9 gene-addition study (cards C01 and C04).

   Available evidence

   WWOX-system, read beyond the abstract: the high-dose arm reaches a survival plateau to day 300 and lowers spike-wave discharges to a wild-type-indistinguishable level; weight is ordered against wild type and the untreated knockout (equal to the high dose at P30 only), and glycaemia is NOT shown to be prevented: at P10 the treated arms are not significantly different from the untreated knockout (high dose about 113, low dose about 105, untreated knockout about 95; Figure 3E, card C04 DR-AAV-GLU10-HD and DR-AAV-GLU10-LD), at P20 the untreated-knockout comparator is absent because those animals are dead (Figure 3F, DR-AAV-GLU20-LD and DR-AAV-GLU20-HD, direction UNKNOWN) and the P20 result is dose-graded against wild type only, and by P30 the dose separation has closed (Figure 3G); the paper's own text states that both doses partially corrected hypoglycaemia, while the panel does not support it (card C04 records this: the treated arms are not significantly different from the untreated knockout at P10); the glucose result depends on age, dose and comparison, and a non-significant difference at P10 is not evidence of no benefit. Every treated arm was injected in the stated neonatal window.

   Development feasibility

   Established model and established vector; the added arm is a timing change, not a new assay.

   Novelty, within the verified scope

   Queried: the two gene-addition cards (C01, C04) and their assertion rows; the deepdive manifest of the source; the work-order statements for this need. Checked for a post-onset arm and found none. Not a statement about the literature beyond the sources this module has read. Extended on 2026-11-05/06 by two bounded external searches (batch_sci_20261105, r1 and r2 query logs, 66 queries in all): conditional, antisense and vector restoration after onset in Mecp2, Ube3a, Fmr1, Tsc1, Pten, Syngap1, Cdkl5, Shank3, Tcf4, Scn1a, Scn1b, Kcnt1, Scn8a and leukodystrophy models. The earlier of the two searches was anchored on gene names and could not reach enzyme, connexin, tubulin or astrocyte-transporter leukodystrophies; an independent review found post-onset white-matter arms there. Bounded searches, not a statement that none exists.

   Reason for the move

   position unchanged from IT20 (rank 1): No IT21 input changes the position. The row text was corrected in IT21 P1 (glycaemia is not called prevented).

   Comparators

   No comparator in the files consulted for this investigation.

2. 1 (tie) Integrated-stress-response modulators (a class; no single compound) new hypothesis generated by LEGEND measurement of a premise, without a compound#

   Question

   Before any ISR-attenuating compound can be ordered: in a WWOX-deficient HUMAN neural system, does the ISR run high or low?

   Reason for the priority

   The need's direction is unresolved and the two available readouts disagree by design, so no compound on this axis can be qualified either way. Resolving the sign is the cheapest act that converts the largest blocked slice of the catalogue into an orderable or a refused question; leaving it unresolved keeps that slice invisible rather than negative.

   Solidity of the evidence (depth of reading)

   E3 - card reviewed with assertions; no WWOX-side result read beyond abstract

   Decisive uncertainty

   Which sign the ISR actually has at protein level in a WWOX-deficient human neural cell, and in which cell type. The IT28 reading of the transferable literature adds that the sign of a lever depends on which node is perturbed and on the stress class (chronic misfolded cargo versus a locked eIF2B lesion versus transient external stress), and that RNA target sets can disagree with eIF2-alpha phosphorylation in both directions. Until the protein-level sign is measured, an ISR-attenuating and an ISR-potentiating compound remain equally defensible.

   Next test

   ATF4 protein, total and phosphorylated eIF2-alpha and per-cell translation (nascent-protein labelling) in a WWOX-deficient human neural line against an isogenic control, at one differentiation stage and by cell type, no compound involved. Only if the sign is high, a pre-registered panel: an eIF2B activator and kinase-level loss against genetic PPP1R15A loss, with a second-stressor arm and a functional endpoint. IT30 (design hypothesis only, from the residues of CH-IT28-ISR-ROLE read in wo06_it30_20261106 and independently checked): keep a network-activity or seizure-relevant readout beside the protein-level sign; if a potentiation arm is tried, make it genetic (a PPP1R15B knockdown) or chemical with measured p-eIF2alpha and an alpha-2 control (guanabenz acts on alpha-2 receptors and Sephin1 and guanabenz were not shown to inhibit PPP1R15A-PP1); the only mammalian seizure-endpoint precedent for a potentiator is two studies of one eIF2-alpha phosphatase inhibitor with opposite signs, so no arm is promoted. IT35 (gsk3_eif2b_20261107, audited): add phosphorylation of eIF2B epsilon at the GSK3 site, eIF2B exchange activity and per-cell translation beside phospho-eIF2 alpha and ATF4, and include a phospho-eIF2 alpha-independent arm; a GSK3 route is not read as equivalent to phospho-eIF2 alpha inhibition; no arm is promoted and no compound is justified by this package.

   Scientific details of this row

   Origin of the proposal

   Generated by LEGEND from the qualified need path: the direction of the integrated stress response is unresolved (IT20).

   Available evidence

   WWOX-system transcript-level only and contradictory: the knockout organoid shows the ISR target set lower, one shRNA library per arm shows it higher, and the designs are not comparable. No ATF4 protein and no eIF2-alpha phosphorylation measurement exists in the sources read. [IT39] Transfer context only: two further ISR reports (VWM line, myelin readouts read in one) do not test WWOX; the biochemical report used as the bridge runs against its use; for human neural cells the absence is 'none found in two PubMed strings', the fly datum is conditional on neuronal Abeta42, and a non-neural WWOX paper links PERK and IRE1.

   Development feasibility

   A protein measurement on material the module already reasons about; no new model and no compound needed.

   Novelty, within the verified scope

   Queried: this need's eleven observation rows and their readout levels; the ISR card and its assertions; the search layer over the eight ISR targets (78 associations, 56 families, none direction-coherent because the need direction is unresolved). The gap is a missing measurement in the sources this module has read, not a claim about the ISR literature.

   Reason for the move

   position unchanged from IT20 (rank 1): IT28 challenge CH-IT28-ISR-ROLE (batches on the neuronal and the glial/myelin literature) changes what the test must measure and which result would discriminate, not the position.

   Comparators

   No comparator in the files consulted for this investigation.

   Conditional physiological hypotheses

   1 conditional hypotheses, with alternatives and what would strengthen or weaken them

   These are questions stated with their conditions, not acquired data: no compound is proposed here and no gene-to-gene link is inferred. The absence of an experiment is not demonstrated harm.

   • ISRIB: eIF2B activator that binds the beta-delta interface and stabilises the decamer; it offsets phospho-eIF2 inhibition and does not stop… (ISR1-INV02; exploratory hypothesis)

     What is observed

     None for ISRIB or eIF2B in a WWOX system (a repository record (not included in this package) CARD-A1, wwox_starting_observation). INV-02 evidence_available: WWOX-system transcript-level only and contradictory: the knockout organoid shows the ISR target set lower, one shRNA library per arm shows it higher, and the designs are not comparable; no ATF4 protein and no eIF2-alpha phosphorylation measurement exists in the sources read (a repository record (not included in this package), INV-02).

     What is hypothesised

     Two readings of a translation difference in WWOX-deficient cells: (1) phospho-eIF2alpha is the brake (ISR route); (2) if not, a GSK3-site brake exists (CARD-A1 inferred_steps). Within reading (2) the sources record further branches: GSK3-other through the mTOR branch, or an unnamed substrate (a repository record (not included in this package) step 3b). Counter-evidence on ISRIB itself: an effect at basal state in some sources and none in others, no ATF4 change in rat cortical cultures, no benefit in one toxic-protein model (CARD-A1 main_counterevidence). A probe response separates (1) from (2) only if the engagement control works and phospho-eIF2alpha is measured.. Exploratory: the package states no therapeutic direction. Unresolved premise: The sign of the ISR at protein level in a WWOX-deficient human neural cell is unmeasured (INV-02 decisive_uncertainty); until it is, an attenuating and a potentiating compound remain equally defensible. Step 0 must first show that the readouts and an engagement control (eIF2B assembly shift by sedimentation, not ATF4 in neurons) can be read in the…

     Alternatives that remain

     Alternatives the package states: Two readings of a translation difference in WWOX-deficient cells: (1) phospho-eIF2alpha is the brake (ISR route); (2) if not, a GSK3-site brake exists (CARD-A1 inferred_steps). Within reading (2) the sources record further branches: GSK3-other through the mTOR branch, or an unnamed substrate (a repository record (not included in this package) step 3b). Counter-evidence on ISRIB itself: an effect at basal state in some sources and none in others, no ATF4 change in rat cortical cultures, no benefit in one toxic-protein model (CARD-A1 main_counterevidence). A probe response separates (1) from (2) only if the engagement control works and phospho-eIF2alpha is measured. || Other rows of the package, not classified by it as rivals, complementary or compatible: ISR2: Only if the protein-level sign is high, a pre-registered panel (INV-02 next_test): an eIF2B activator and kinase-level loss against… || ISR3: INV-02 next_test (IT30 design hypothesis): if a potentiation arm is tried, make it genetic (a PPP1R15B knockdown) or chemical with…

     What would strengthen it

     none claimed: the probe gives information, it is not assumed to rescue (CARD-A1 hypothesised_benefit) Measure: a repository record (not included in this package) step 1 (condition: step 0 passed): (a) phospho-eIF2alpha raised and ISRIB restores translation: ISR route in this preparation, go to step 4, GSK3 branch deprioritised but not refuted; (b) phospho-eIF2alpha not raised and translation lowered: go to step 2 (GSK3 contribution); (c) translation not lowered: stop, change stage or cell type before concluding; (d) engagement control fails: uninformative, repeat step 0.

     What would weaken it

     Weakened if phospho-eIF2alpha is not raised and ISRIB with a working engagement control gives full translation restoration (then ISRIB acts by a route not described in the sources read) (CARD-A1 result_that_would_weaken_it); repeat step 0 if the engagement control fails.

3. 2 (tie) Blockade of the T-type calcium current by a proposed pair of approved agents, to verify isoform reach against the carded T-type agents new hypothesis generated by LEGEND thread with a result already in a card#

   Question

   Which T-type isoform carries the suppression of the provoked seizure phase, and does the effect depend on the genotype?

   Reason for the priority

   The module's one T-type result is complex-or-group level with no isoform datum and no drug-by-genotype interaction test, and its own need file names both gaps. Two approved compounds are proposed as a pair that MAY differ in isoform reach; the pair is a proposal to verify, not a documented discriminating pair: the catalogue difference between them (one T-type target row for the first agent, three for the second) comes from annotation completeness, not from a documented functional selectivity, because in the catalogue both are tagged selectivity Not Determined and no fitted potency on the three isoforms is carried for the first agent (ChEMBL holds no calcium-channel activity for it), while the second has Kd values on all three only from synthesis papers that used it as a reference, in different assays and conditions. Absence of a subtype in a catalogue is not inactivity. Both agents also act on other targets (L-type channels for one, receptors beyond the channel family for the other), so the potency measures in one assay and species, at the concentration reached in the animal, are the first step.

   Solidity of the evidence (depth of reading)

   E3 - card reviewed with assertions; no WWOX-side result read beyond abstract

   Decisive uncertainty

   Whether the carded effect is a T-type effect at all, which isoform carries it, and whether it is genotype-dependent. Whether the two agents really differ in isoform reach at the concentrations used is itself unmeasured. The same effect from two compounds is compatible with a common target (the T-type class) and does not by itself weaken the T-type attribution; it only fails to discriminate isoforms. What would weaken the attribution is no effect of either agent at on-target exposure, an effect that tracks off-target activity rather than T-type potency, or an equal effect in wild type (no genotype dependence). [IT36] Z944 was measured on Cav3.2 only in the source read (human clone, HEK293T, whole cell); Cav3.1 and Cav3.3 values and mouse brain exposure are not obtained; the blocker class has not discriminated between isoforms.

   Next test

   The published provoked-seizure design with a drug-by-genotype interaction term and the proposed pair beside the carded agent, after the potency of each agent on the three isoforms has been qualified in one assay and species; identical provocation and scoring.

   Scientific details of this row

   Origin of the proposal

   Reached from the T-type calcium need through the catalogue path; the proposed pair is a proposal to verify (IT20).

   Available evidence

   WWOX-system: a T-type blocker suppressed the late phase of the provoked seizure score in the null mouse, one experiment, genotype specificity not established by an interaction. Catalogue: direction-coherent blocker actions on CACNA1G, CACNA1H and CACNA1I, at group and single-protein level, from curated mechanism rows. [IT39] The panel of 'structurally unrelated' antagonists is not independent: two of the listed antagonists are the same molecule under different names (one of them is Z944), and Z944 and TTA-P2 share a core chemical scaffold.

   Development feasibility

   Published assay, approved compounds, three arms added to an existing design.

   Novelty, within the verified scope

   Queried: the search layer over CACNA1G, CACNA1H and CACNA1I (190 associations, 66 families); the T-type card and the ethosuximide and apinocaltamide units; the uncarded-items table; the IT9 to IT11 qualification records; a name scan of every card and ranking table for each proposed compound. Both proposals are absent from that WWOX material and are members of a class the module already holds, so this is a proposed probe set to verify, not a new biological relation.

   Reason for the move

   position unchanged from IT20 (rank 2): No IT21 input changes the position. The row text was corrected in IT21 P1 (the pair is a proposal to verify, not a documented discriminating pair). IT29: in other models the sign of several channel-agent classes flips with the lesion class; a discriminating challenge in the WWOX slice preparation would separate an inhibition-loss-like from an excitatory-gain-like network.

   Comparators

   3 functional-outcome, 5 mechanism-separating (compound names are not in the shareable edition)

4. 2 (tie) Ethosuximide (see its card) and the carded T-type thread thread already known in WWOX thread with a result already in a card#

   Question

   Is the carded suppression specific to WWOX loss, or does the same agent act equally in wild type?

   Reason for the priority

   The position of this thread rests on a single laboratory and a single paper in which the comparison drug acted better, and the specificity flag is an unknown rather than a clean reading. The thread is already ordered, so the result changes a standing position rather than adding a new one.

   Solidity of the evidence (depth of reading)

   E4 - card reviewed, 1 WWOX-side assertion(s) read beyond abstract

   Decisive uncertainty

   Whether an interaction term survives; a wild-type effect of the same size would move this thread out of its group without any new compound being involved.

   Next test

   The same provoked-seizure design with both genotypes treated and an interaction term, powered on the published effect size.

   Scientific details of this row

   Origin of the proposal

   Carded thread: the T-type suppression of the provoked seizure phase in the null mouse.

   Available evidence

   WWOX-system with a specificity flag: one paper, one laboratory, no drug-by-genotype test, and a wild-type arm whose variability is comparable to the effect.

   Development feasibility

   The assay exists and the compound is carded; this is the control arm the original design omitted.

   Novelty, within the verified scope

   Queried: the ethosuximide and apinocaltamide units, their assertion rows and caveat rows, and the per-need position table. The missing interaction test is recorded in this module's own caveats, not inferred from absence.

   Reason for the move

   position unchanged from IT20 (rank 2): No IT21 input changes the position. IT29: epilepsy-model screens in other genotypes give no corroboration for this agent (absence of corroboration is not refutation); the position rests on the single pup experiment.

   Comparators

   No comparator in the files consulted for this investigation.

5. 3 (tie) NMDAR antagonism (the carded tool compound and the clinically used class) thread already known in WWOX thread with a result already in a card#

   Question

   Does a clinically used NMDAR antagonist reproduce the burst suppression obtained with the tool compound, at a stated extracellular magnesium concentration?

   Reason for the priority

   This is the only axis in the module with a measured WWOX-system drug readout, and it was obtained with a full competitive block by a tool compound. Whether a partial block reproduces it is the step between a perturbation result and an orderable question; the magnesium dependence makes the result uninterpretable unless the bath is reported.

   Solidity of the evidence (depth of reading)

   E4 - card reviewed, 1 WWOX-side assertion(s) read beyond abstract; class pharmacology table (IT7 P5)

   Decisive uncertainty

   Whether a partial, clinically achievable block suffices, and whether the carded effect is a subunit effect or a pore-block effect. A pore blocker's potency depends on voltage and on magnesium, so the unreported bath is itself the obstacle. [IT36] A GluN2B NAM effect would show participation of GluN2B-containing receptors in sustaining bursts, not that the lesion lies in GluN2B; it needs a target-specific control (matched-current d-APV). [IT40] B5 Radiprodil is GluN2B-selective in oocytes (the selectivity sentence rests on an unread supplementary figure) with 45 percent maximum inhibition of GluN2A/GluN2B triheteromers and very slow washout; its use or activity dependence was not tested; its slice precedent is CA1 at P17-P22 with bath magnesium stated as 1.5 mM in the Methods and 0.2 mM in the legends. TCN-201 acts on recombinant GluN2A only, potency set by glycine, maximum inhibition 73-81 percent, so a TCN-201 arm needs a measured slice co-agonist level.

   Next test

   Repeat the slice burst assay with the magnesium concentration reported, a d-APV arm and a GluN2B NAM arm, each at matched measured NMDAR-current reduction (one ifenprodil-site GluN2B NAM; a second NAM of the same site controls chemotype only, and only if its own maximum on triheteromers is known (ifenprodil 32 percent, radiprodil 45 percent in rat: each molecule has its own); compare with d-APV only up to the NAM's measured removable share; measure the d-APV abolition threshold x and the NAM removable fraction first), vehicle time-matched, animal as the unit; memantine only afterwards, with the magnesium concentration and holding voltage stated, beside the tool compound, and the untreated same-genotype comparator. [IT40] B6 Report bath magnesium, K+, temperature and blockers actually present as measured values; use a within-slice paired design; add an AMPA/kainate block arm verified by the fast EPSC (abolition alone is not informative, because loss of magnesium relief would abolish an NMDAR-regenerative burst too); add a matched-effect control read on at least two metrics with a receptor-agnostic release-reducing control; add a GABA-A arm because the sIPSC amplitude is halved; state the order of application and washout as part of the design. [SCI42] Report bath pH (and the buffer system) as a measured value with magnesium and potassium; read the percent of evoked synaptic charge removed for every drug so the matched drive-reduction control can be matched; a pH-dependent GluN2B tool is considered only after the tissue pH is measured, with the ifenprodil-site tool and d-APV at the same two pH values.

   Compounds or classes involved (research order)

   • control Full competitive NMDA-receptor block, the reference control
   • probe Ifenprodil-site GluN2B negative modulator, used as a tool
   • still unresolved AMPA/kainate block arm: a pair of two mechanisms
   • known candidate, requalified Supply-side GABA lever: block of GABA reuptake, a different question from receptor potentiation
   • known candidate, requalified Clinical-stage GluN2B negative modulator with an assembly-dependent maximum
   • still unresolved Human-exposure GluN2B negative modulator; site, maximum effect and triheteromer behaviour not measured in any source reached
   • probe pH-dependent GluN2B modulator: a pH-linked comparison, subordinate to a measured tissue pH; not a candidate
   • still unresolved GABA-transaminase block: affinities of irreversible inhibitors are not a potency ranking
   • still unresolved Class recorded on a GluN1/GluN2B assembly record, no maximum effect
   • weakened or not to be transferred The registry GluN2B row is a GluN2B measurement (113 micromolar) stored as nanomolar; DQP-1105 prefers GluN2C and GluN2D: excluded from the GluN2B comparison
   • probe A second ifenprodil-site tool: primary potency and site are read, no maximum inhibition and no triheteromer figure

   Scientific details of this row

   Origin of the proposal

   Carded axis: burst suppression by NMDAR block in the slice.

   Available evidence

   WWOX-system: blocking NMDAR abolished the burst in the WWOX-deficient slice, one drug, one preparation. The recording bath magnesium is not stated in the source, and the incubation solution already contains magnesium. [IT36] Memantine block of extrasynaptic current was about 1.9-fold that of synaptic in one autaptic design; a later paper in culture found little extrasynaptic selectivity at 2 uM; preference is not exclusion and the minutes-scale premise is untested at WWOX burst durations. [IT36] burst duration median 0.66 s (Supplementary Table S1), intermediate between a synaptic pulse and the minutes-scale activation of the memantine premise; the premise is untested at this scale. [IT39] Selfotel: the head-injury trial was stopped after a signal in a stroke study, and a later analysis did not confirm excess mortality in head injury (abstract level). Perzinfotel is NR2A-preferring. The carbenoxolone half of the gap-junction logic carries its own caveat (it is not specific to gap junctions). [ADV-004 2026-11-14] The WWOX-side evidence of this row comes from the same Wwox slice paper whose unit-of-analysis limits are recorded under INV-10 [IT47 SCI47]: the change in inhibitory-current frequency rests on a distribution test on pooled events (no per-cell or per-animal test), the excitatory-current pipette holds about 12 mM chloride so isolation at -70 mV is not chloride-free, and burst frequency differs between the Results and the supplementary table (a repository record (not included in this package)).

   Development feasibility

   The slice preparation exists; the change is a reported bath and one added arm. [IT41] Decision: whether a GluN2B-directed comparison is the better question (needs the d-APV abolition threshold x and one modulator's removable fraction, read at the burst). Possible in silico or from public sources: a primary source for the site and maximum effect of a second modulator; qualification of the AMPA tools. New measure: isolated NMDA-receptor current, x, removable fraction, measured bath magnesium and pH. Reopen if: a second modulator is shown to have a different ceiling at the same site, or the AMPA-tool qualification is delivered. Priority: first among the compound-facing investigations, because it is the cheapest measurement that decides whether the other arms are asked.

   Novelty, within the verified scope

   Queried: this need's eight observation rows, the two NMDAR cards and their assertions, and the search layer over the eight receptor-subunit targets (709 associations, 187 families). The module's own need file records that the clinically used agents were not found to have been given to a WWOX system in the sources read.

   Reason for the move

   position unchanged from IT20 (rank 3): No IT21 input changes the position.

   Comparators

   4 functional-outcome, 7 mechanism-separating (compound names are not in the shareable edition)

   Conditional physiological hypotheses

   3 conditional hypotheses, with alternatives and what would strengthen or weaken them

   These are questions stated with their conditions, not acquired data: no compound is proposed here and no gene-to-gene link is inferred. The absence of an experiment is not demonstrated harm.

   • Does the burst depend on AMPA-receptor drive? (BH1; exploratory hypothesis)

     What is observed

     In the neuron-specific knockout slice an NMDA-receptor antagonist abolishes the spontaneous bursts. No AMPA/kainate antagonist has been applied; a shift of the knockout excitatory-current amplitude fits an AMPA-type quantum only by inference.

     What is hypothesised

     The burst may also need AMPA-receptor drive. This is exploratory; the premise is not established.

     Alternatives that remain

     NMDA-receptor-regenerative drive (GluN2B- or GluN2A-weighted), a general loss of excitatory drive, and a recording-state artefact. Kainate receptors are not separated from AMPA receptors by the usual tools.

     What would strengthen it

     A verified AMPA/kainate block that abolishes the bursts at less total charge removed than the NMDA-receptor antagonist needs would put AMPA receptors first in the next question.

     What would weaken it

     Bursts that persist under a verified, pre-specified AMPA/kainate block weaken this reading. Abolition alone does not separate AMPA drive from NMDA-receptor-regenerative drive, because losing the depolarisation also removes magnesium relief.

   • Is the NMDA-receptor dependence total receptor gain or GluN2B-weighted? (BH2; exploratory hypothesis)

     What is observed

     The burst needs NMDA-receptor activity; about 20 percent of the delayed evoked response remains against the wild-type floor under the antagonist, which is not evidence of incomplete block.

     What is hypothesised

     A GluN2B-selective negative modulator may remove the bursts at a smaller measured fraction of the isolated NMDA-receptor current than the full antagonist needs. Exploratory.

     Alternatives that remain

     Total NMDA-receptor reduction, a GluN2A-weighted component, a general loss of drive, and a bath-state (magnesium, potassium) effect.

     What would strengthen it

     A curve shifted to the left of the full-antagonist curve, with a measured sensitive fraction, makes the GluN2B-directed comparison the better question.

     What would weaken it

     Coincident curves make total NMDA-receptor reduction (with the bath magnesium and voltage) the question; a plateau below abolition moves it to GluN2A-containing receptors. None of this can come from other preparations.

   • Can the gap-junction tool's burst fall be read as a gap-junction result? (BH3; exploratory hypothesis)

     What is observed

     A gap-junction-blocking tool lowered the burst rate in the knockout slice and did not recover. In culture the same tool lowers AMPA and NMDA currents, broadens spikes and lowers input resistance.

     What is hypothesised

     The fall may come from effects on synaptic currents and excitability rather than from coupling. No datum in the sources read supports the gap-junction half. Exploratory.

     Alternatives that remain

     Reduced release or altered excitability; direct NMDA-receptor channel block (neither shown nor excluded); neuronal electrical coupling; pannexin-1; run-down or a recording artefact.

     What would strengthen it

     A residual fall beyond a control matched to the same measured current reduction, together with a loss of coupling, keeps the connexin question open.

     What would weaken it

     If a matched non-gap-junction control gives the same fall, the gap-junction attribution is not supported; a pannexin-1 blocker with shown engagement that lowers bursts would support pannexin-1.

6. 3 (tie) Pro-myelinating agents (a class; no single compound) new hypothesis generated by LEGEND measurement of a premise, without a compound#

   Question

   Is there an axonal signal left for a pro-myelinating agent to act on, before any such agent is ordered?

   Why it matters, what we know, next step

   Reason for the priority

   This need carries 318 catalogue associations and not one of them is direction-coherent: a restore-type need has no action vocabulary in the catalogue to match, so the slice looks empty for a structural reason, not a biological one. The premise question is cheaper than any compound test and decides whether the axis is a target or a consequence.

   Solidity of the evidence (depth of reading)

   E2 - card reviewed, abstention (premise named, no test designed)

   Decisive uncertainty

   Whether the neuron still emits the activity-dependent signal that a pro-myelinating agent would amplify. If it does not, the whole class is not a lever here regardless of its pharmacology. IT29: even where an intrinsic route works, whether it acts when the lesion is neuronal is untested. [IT36] no evidence of early hyperactivity exists for the neuronal-deletion model before P13; activity effects on markers in the neonate (P0-P8) do not speak to the burst window. [IT36] The cue from the neuron to the precursor is not sourced; the neuregulin-ErbB version is counter-evidenced in vivo to postnatal day 11.

   Next test

   Measure the axonal activity-dependent myelination signal in the WWOX-deficient system against an isogenic control, before any compound; then decide whether the class is orderable. IT29: add an arm in Wwox-null precursor monoculture or the Wwox-null neuron co-culture, with a lipid-synthesis comparator beside the muscarinic one, crossed with a neuronal-activity manipulation. [IT36] existing co-culture separates neither contact or substrate integrity nor a soluble product; peripheral seizure-free DRG neurons; add a fixed-neuron arm and a factor arm. [SCI50 update] The conditioned-medium arm needs a fresh-medium baseline, a mixing arm, neuron purity, a dilution series, the B27/T3 caveat, a positive control and a detection limit taken from the monoculture's own variance; neuron-secreted CTGF is reported as necessary and sufficient to block OPC maturation in neuron-specific Tsc models (the authors' in vitro claim; no Wwox datum). Variant of step X3 behind gate 0 (T3 and T4 in the medium, blind reproduction of the null-neuron deficit, an OPC window with M1 tone): an M1-selective muscarinic antagonist and an EBP inhibitor without M1 affinity, crossed with neuron genotype; null contrasts are read only when positive; a persisting gap is predicted both by a neuronal ceiling on sheath structure and by an OPC-intrinsic block. The M1/EBP premise is unresolved. IPOTESI; the Wwox-CTGF link is unsupported. [SCI51 update] The M1-selective arm is a design role, not a sourced reagent: the reagent named earlier has no public structure (a licence agreement withheld it; a correction and an expression of concern exist), whether it can be obtained from its developer is not stated in anything read, and its EBP activity is unmeasured in every source reached (unmeasured is not inactive). A role is filled only when a reagent of disclosed structure has its EBP activity measured in the OPC preparation (readout of a pathway intermediate or enzymatic assay) beside an EBP inhibitor; M1 tone is measured at a controlled plating density with an ACh add-back; the pair is read only after these gates, at submaximal doses with a T3 headroom arm. No M1-selective antagonist has an EBP datum in the sources reached (two antimuscarinic drugs have positive EBP data; four unnamed antimuscarinics were EBP-negative in one OPC line); the CW compounds are M1-low in one single-point assay, not shown M1-clean; the null-baseline occlusion is weaker than first stated but remains. IPOTESI; no reagent is qualified or disqualified.

   Scientific details of this row

   Origin of the proposal

   Generated by LEGEND from the myelination need (IT20).

   Available evidence

   WWOX-system: the myelin deficit is partly prevented when WWOX is restored in neurons in the neonatal window; the treated arms of the source are unquantified, and whether an established deficit re-myelinates is unmeasured. IT29: perturbation evidence from other systems signs several intrinsic precursor-differentiation routes (muscarinic-receptor, lipid-synthesis, thyroid- and RXR-type receptor) in wild-type or injury settings and none in a neuron-driven hypomyelination model; HDAC3 loss is adverse for maturation; an independent screen disagrees with some signs. [IT36] GSK3 sign on oligodendrocyte differentiation is divided (inhibitors raise it in one source, lower it in another after Foxg1 knockdown; genetic loss in mature cells changed survival); recorded as undetermined. [IT39] GSK3 context rows: a GSK3 inhibitor run after Foxg1 knockdown goes the opposite way to the inhibitor-raises-differentiation reports, and a 2026 review holds both directions; the slice paper's finding is neuron-specific wording, not oligodendrocyte-wide. [ADV-001] The muscarinic-receptor route rests on genetic rows (culture; abstract level) with a ceiling caveat; for two antimuscarinic drugs of the class the drug-class support carries a documented EBP-inhibition route whose attribution is unsettled, and EBP activity is unmeasured for the other antimuscarinics named. A profile of the pathway intermediates per compound before target attribution is a design position (INFERENZA); this does not say that EBP explains the muscarinic route, that unmeasured compounds act through EBP, or that M1 is irrelevant. The 216-row signed layer is not committed and no number is reproduced on it.

   Development feasibility

   A measurement in an existing system; no compound and no new model.

   Novelty, within the verified scope

   Queried: this need's three observation rows; the pro-myelinating class unit and its abstention; the search layer over the six myelination targets (546 associations, 313 families, none direction-coherent). The emptiness is a vocabulary mismatch this package measured, not a claim that no pro-myelinating literature exists.

   Reason for the move

   position unchanged from IT20 (rank 3): No IT21 input changes the position. The thyroid-hormone probe row asks the receptor-responsiveness question on the same axis; the two are not merged because one measures the neuron's signal and the other the precursor cell's response. IT29 (phenotype-first table): position unchanged; the added evidence is transferred from non-WWOX systems and conditional.

   Comparators

   No comparator in the files consulted for this investigation.

7. 4 (tie) Dichloroacetate (see its card) and inhibition of the pyruvate dehydrogenase kinases thread already known in WWOX measurement of a premise, without a compound#

   Question

   Is there a PDH block in a WWOX-deficient human neural cell at all, and does its direction match the one this lever assumes?

   Why it matters, what we know, next step

   Reason for the priority

   The chain that motivates this lever predicts a PDH block, and the two available data point the other way: the kinase transcript is down and the one human activity measurement shows no difference. Ordering the compound before the premise is measured is how a plausible chain becomes a wasted experiment.

   Solidity of the evidence (depth of reading)

   E3 - card reviewed with assertions; no WWOX-side result read beyond abstract; class pharmacology table (IT7 P5)

   Decisive uncertainty

   Whether PDH flux is actually reduced. The lever's direction is derived from the chain, not from a measurement, and the available signs contradict it. [IT36] An energy-to-myelin link (glucose or lactate supply to the oligodendrocyte lineage) is a new thread with undetermined sign: WWOX-side lactate data point up, and when glucose falls in the neuron-targeted line is not measured (hypothesis H2 and open question OQ3 of myelin_generative_it36_20261108). [IT43] The signs conflict between systems: the kinase transcript is lower in the organoid, and higher with lactate up and activity unchanged in a non-neural fibroblast line; no neural measurement exists (package metabolic_inv07_20261107). The one randomised human trial of the reference, in a mitochondrial disease, was stopped early for peripheral nerve toxicity (abstract level), a safety limit to read before any arm is proposed. [SCI44] Lowering PDK is also a risk arm: documented harm directions (6 Hz proconvulsant at abstract level, only the combination of DCA and pyruvate helped in rat seizure, peripheral myelin proteins fell, neuropathy in the one randomised trial) and a lactate-supply role make the net direction by class unknown.

   Next test

   PDH activity and flux in a WWOX-deficient human neural cell against an isogenic control, with the glycolytic readout measured in the same material. [IT36] ECAR is not a lactate readout without an oxygen correction; measure lactate and pyruvate directly with redox state. [IT43] The isogenic pair and the oxygen series of INV-19 can carry this measurement: amount and phosphorylation of the step at all three sites, activity before and after dephosphorylation, and the four kinase transcripts, read next to BNIP3 from the same lysates. Reading (IPOTESI): BNIP3, kinase and phosphorylation all up supports a HIF-linked restraint and keeps the kinase lever conditional on it; BNIP3 up with the step unrestrained leaves the lever down while INV-19 stands; neither up points to the culture or oxygen condition. The compound arms follow only after a restraint is measured. [SCI44] Read the step on marker-resolved or sorted material; the dominant kinase RNA differs by class and species (mouse neurons PDK2; human neurons PDK3; astrocytes PDK4 in both species; human oligodendrocytes PDK4): include PDK2-4 transcript and protein per class, a genetic knockdown of the dominant isoform before any compound, and the scaffold condition of any tool that acts through the enzyme-complex scaffold. A restraint measured in one class does not transfer to another.

   Compounds or classes involved (research order)

   • known candidate, requalified Lowering the kinases that restrain the pyruvate dehydrogenase step: a carded lever with typed, per-isoform rows, conditional on an unmeasured premise; the first probe depends on the dominant kinase of the cell class and on the assay scaffold
   • probe A lactate-side probe: the sign depends on the cell compartment
   • probe A probe at the opposite end of the node and a fuel control, to separate a dehydrogenase defect from a supply defect

   Scientific details of this row

   Origin of the proposal

   Carded chain: Dichloroacetate (see its card) and inhibition of the pyruvate dehydrogenase kinases on the pyruvate axis.

   Available evidence

   WWOX-system: a glycolytic signature; the kinase transcript lower in the organoid; no difference in the one human PDH activity measurement, which is a non-neural fibroblast pool with partial WWOX depletion (n not stated) and says nothing about neurons. The kinase-to-PDH link has no WWOX observation. [IT39] A ketogenic-diet response is reported in 3 of 5 WOREE individuals (lactate 2.4 to 3.3 mmol/L in one individual); the Pdk1 transcript and serum-lactate datum in the Wwox-null MEF and mouse paper is non-neural and measures no PDH.

   Development feasibility

   An established assay on material the module already uses.

   Novelty, within the verified scope

   Queried: this need's ten observation rows and their readout levels; the dichloroacetate and HIF-axis units; the search layer over the eight axis targets (209 associations, 180 families, none direction-coherent because the need direction is unresolved).

   Reason for the move

   position unchanged from IT20 (rank 4): No IT21 input changes the position. One IT21 rejection (an inhibitor of class I histone deacetylases proposed to lower SLC2A1) touches the same glycolysis theme and is listed under what the generator rules out.

   Comparators

   No comparator in the files consulted for this investigation.

8. 4 (tie) The NRF2 (KEAP1) axis, genetic arm before any compound thread already known in WWOX measurement of a premise, without a compound#

   Question

   Does lowering KEAP1 or raising NRF2 change a functional readout in a WWOX-deficient neural cell, in the direction the two conditional arms of this need assume?

   Why it matters, what we know, next step

   Reason for the priority

   Both arms of this need are explicitly conditional and neither is established, and the module has already refused to read agreement between two electrophiles as evidence of pathway dependence. A genetic arm answers the dependence question that no compound pair can answer.

   Solidity of the evidence (depth of reading)

   E3 - card reviewed with assertions; no WWOX-side result read beyond abstract

   Decisive uncertainty

   Whether the axis is engaged at all in a WWOX-deficient neural cell, and whether engaging it helps or harms. No compound in the record acts in such a cell. IT22: the AhR-target transcript is higher in the knockout while NFE2L2 is lower, so engagement of the axis is not even directionally clear. [IT45] Induction of one enzyme can occur without NRF2 (reported for compounds that do not bind KEAP1), a Kelch-domain binding assay cannot see covalent modifiers, and the one catalogued NRF2-lowering row lowers protein synthesis in general: none of these is a control for an NRF2-independent benefit. [IT46 SCI46] Where an NRF2 loss was verified (primary rat microglia) the binder's effect was lost and KEAP1 loss occluded it; in a human neuronal line a partial loss left the effect unchanged (uninterpretable); one human neuronal use on an already NRF2-high background moved ferroptosis markers in the harmful direction; the source's own siRNA arms are categorical labels in the database copy. Nothing is measured in a WWOX-deficient neural system. Source: a repository record (not included in this package).

   Next test

   Genetic KEAP1 lowering with a functional readout in the WWOX-deficient neural system, isogenic control, before any electrophile is compared. [IT36] precedent for a non-cell-autonomous (astrocyte) route exists in another system with an Nrf2-null control and rescue; KEAP1 knockdown needs an NFE2L2-loss cross; composition is not astrocyte-mediated benefit. [IT46 audit] The covalent and the non-covalent compound are not compared at equal concentration and time (the electrophile is consumed by thiols and persists after washout): a phase 0 matches exposure on a proximal NRF2 reading (concentration-response, time-course, cytotoxicity, glutathione, washout) first; at least two independent NFE2L2-knockdown reagents with measured efficiency; an NRF2-independent control gene set beside the target set; a WWOX genotype factor, since the pair speaks to the NRF2 pathway and not yet to the WWOX hypotheses; neuron and astrocyte cultures have no network, so a network outcome is a second phase. Source: a repository record (not included in this package); a repository record (not included in this package). [IT46 SCI46] Reagent pair named (SCI46): the non-covalent KEAP1 Kelch binder (a catalogue tool probe) against a covalent comparator, in one neural cell class at a time. X0: nuclear NRF2, two target genes (G6PD may be one), NQO1 enzyme activity and a risk panel (glutathione, GPX4, lipid peroxidation); the same lysates serve INV-07 step 1 and INV-19 (a shared preparation, not a shared cause). X1: a graded occupancy ladder matched on on-target response, not on concentration. X2: an adequacy criterion for NFE2L2 and KEAP1 loss fixed before the experiment, a KEAP1-loss occlusion arm, and a panel or array whose survivors after loss are read beyond NQO1. X3: neuron-only and astrocyte-only loss. Viability is read beside a function endpoint. Source: a repository record (not included in this package), a repository record (not included in this package), a repository record (not included in this package) (NH1-NH4). [IT46 SCI46q] Readouts by cell class (SCI46 queue item, RNA class means, mouse P7 and P17 oligodendrocyte-lineage, one human neuron sample; not protein or activity): a neuron-readable target panel (Gclm, Txnrd1, Gpx4, G6pdx, Srxn1, Gclc) beside nuclear NRF2; NQO1 and HMOX1 as induced-state readouts with the per-class baseline (NQO1 is at the floor in mouse neurons, 0.84 FPKM, and low in astrocytes, 3.02; HMOX1 is 2.43 in neurons, 14.85 in astrocytes and 260.42 in microglia); HMOX1, NFE2L2 and SLC7A11 are astrocyte-biased, so a bulk transcript readout can move with cell composition alone; the HCAR2 route of a fumarate needs a microglia-like cell (Hcar2 18.12 FPKM in microglia against 0.10 in neurons and astrocytes). Source: a repository record (not included in this package). [IT47 reading limits] RNA class means do not say what a cell can do: a low value is not an absent function, a gene readable in neurons is not thereby an NRF2-specific readout (other regulators act on each), and KEAP1 RNA in a class does not show that the protein is available to a binder. The held RNA reads Hcar2 only in microglia, so a culture without microglia is not expected to test the HCAR2 route of a fumarate unless Hcar2 protein is shown in the cells used. [SCI48] Step X2 adds: loss verified as protein and mRNA at the endpoint time (not an allele PCR); the electrophile inside the loss arm beside the binder; loss-alone and reagent-alone arms under the same challenge; two electrophile exposures matched on a proximal on-target response (one target gene is not enough). Step X1 adds the inactive analogue of the binder as a control. IPOTESI; does not speak to WWOX; conditional on X0.

   Compounds or classes involved (research order)

   • still unresolved Raising NRF2 activity: a covalent KEAP1 modifier and a non-covalent KEAP1 binder, told apart by a knockdown of NRF2, not by binding potency

   Scientific details of this row

   Origin of the proposal

   Carded axis: the NRF2/KEAP1 need with two conditional arms.

   Available evidence

   Outside WWOX for the pharmacology; WWOX-system only for the raised reactive-oxygen readouts, whose sign differs by context between a fly and late-passage fibroblasts and does not reconcile. IT22: the organoid NFE2L2 transcript (0.75x) is now a generator input (new question, role unknown, second system failed its control); the generator returns a topical AhR-pathway agent and agents that block EZH2 as conditional rows, and the canonical NRF2 activators only as diagnostic direct hits on the seed gene itself. [IT39] The Src and Fyn chain to NRF2 from an external report rests on a retracted article (21690096); no consumer carries the chain now, its other support is a 2007 paper from the same laboratory line (not independent), and reductive stress is a transfer from cardiac models.

   Development feasibility

   Standard genetic perturbation; the readout is the constraint, not the reagent. [IT45] Decision: whether the genetic NRF2 arm precedes any compound, and which KEAP1 pair separates an NRF2-dependent from an independent benefit. Possible in silico: done for the catalogue (a dose-response activity read reaches catalogued covalent modifiers and a non-covalent binder, each with a primary document, five of eight read only as abstract or database copy). New measure: NRF2 protein, nuclear NRF2 and target-gene activity per compartment in a WWOX-deficient neural system, then each compound with and without NFE2L2 knockdown. Reopen if: the primary text of the abstract-only documents or neural-cell data change a premise.

   Novelty, within the verified scope

   Queried: this need's three observation rows; the NRF2-axis unit and the IT10 and IT11 qualification records on it; the search layer over NFE2L2 and KEAP1 (890 associations, 863 families, none direction-coherent because the need direction is unresolved).

   Reason for the move

   position unchanged from IT20 (rank 4): No IT21 input changes the position. An ALOX15 proposal on the same axis was left inconclusive in IT21 (no neural WWOX measurement). IT22: unchanged position; a new generator input reaches the axis but adds no measurement.

   Comparators

   No comparator in the files consulted for this investigation.

   Conditional physiological hypotheses

   2 conditional hypotheses, with alternatives and what would strengthen or weaken them

   These are questions stated with their conditions, not acquired data: no compound is proposed here and no gene-to-gene link is inferred. The absence of an experiment is not demonstrated harm.

   • Response insufficient for the stress (low NRF2 tone or reserve) (NRF2-H1; exploratory hypothesis)

     What is observed

     Recorded support, not a new observation (it may include the package's own audit remarks): NFE2L2 and NQO1 transcripts lower in W01, NFE2L2 lower in W02; ROS higher in fly and late-passage MEF [Audit: the two ROS supports come from systems with no NRF2 measure.]

     What is hypothesised

     Response insufficient for the stress (low NRF2 tone or reserve). Exploratory: the package states no therapeutic direction.

     Alternatives that remain

     Other rows of the package, not classified by it as rivals, complementary or compatible: H2: Response already induced but functionally insufficient || H3: Benefit mediated by astrocytes or another compartment || H4: Differences between cell types or developmental phases || H5: Compound benefit independent of NRF2 || H6: No benefit or worsening

     What would strengthen it

     Not stated by the package (missing): a falsifiable prediction is not given.

     What would weaken it

     NRF2 nuclear protein, NQO1 activity and ROS before and after a fixed challenge, per genotype and clone

   • Response already induced but functionally insufficient (NRF2-H2; exploratory hypothesis)

     What is observed

     Recorded support, not a new observation (it may include the package's own audit remarks): HMOX1, SRXN1, GSR higher; MEF early redox transcripts higher; P47T Hmox1/Srxn1 up

     What is hypothesised

     Response already induced but functionally insufficient. Exploratory: the package states no therapeutic direction.

     Alternatives that remain

     Other rows of the package, not classified by it as rivals, complementary or compatible: H1: Response insufficient for the stress (low NRF2 tone or reserve) || H3: Benefit mediated by astrocytes or another compartment || H4: Differences between cell types or developmental phases || H5: Compound benefit independent of NRF2 || H6: No benefit or worsening

     What would strengthen it

     Not stated by the package (missing): a falsifiable prediction is not given.

     What would weaken it

     induced target activity plus a functional deficit that further NRF2 raising corrects (or does not)

9. 5 (tie) Lithium thread already known in WWOX thread with a result already in a card#

   Question

   Should this thread stay in the selection, given that its only pharmacology acts against a second need of the same objective?

   Why it matters, what we know, next step

   Reason for the priority

   The thread is carried by one paper from one laboratory, the agent is non-selective, and it pushes the canonical Wnt readout in the direction this module's other need wants lowered. The question now is whether to keep it, not what to measure next with it: that is a decision a single interaction test can settle.

   Solidity of the evidence (depth of reading)

   E4 - card reviewed, 1 WWOX-side assertion(s) read beyond abstract

   Decisive uncertainty

   Whether any part of the carded effect is attributable to the named kinase rather than to the agent's other actions, and whether the opposing push on the second need is tolerable. [IT39] An inositol-depletion arm is not GSK-3-free (GSK-3 alpha inactivated, beta not significantly altered; one paper, one isoform); the S9A cross is not an attribution control; time-gated dosing has no timing source.

   Next test

   A selective inhibitor of the named kinase beside the non-selective agent in the same WWOX-system assay, with the second need's readout measured in the same experiment. IT21 addition: include an arm with the inositol precursor added back, which tests the second action in the same experiment. [IT36] selectivity is biochemical and does not carry to cells (4-6-fold in live cells); attribution needs an isogenic isoform-knockout check per compound or convergent pharmacology. [IT36] in rat organotypic CA1 the GSK3-PI4KIIalpha branch sets synaptic NMDAR current; compare GSK3-route compounds with d-APV at matched NMDAR-current reduction; double GSK3A/B depletion needed. [IT47 package] Hypotheses read by the generator from the GSK3 and eIF2B package (five rows, no direction asserted, no repurposing candidate named): before any compound arm, the baseline of eIF2B epsilon Ser535 phosphorylation (human Ser540) with total GSK3 and replicate lanes in the WWOX-deficient cells against corrected controls; the isoform (beta or alpha) by genetic knock-down; a phospho-resistant epsilon (Ser to Ala) and a catalytically dead epsilon as epistasis, with two limits found by the independent reader: Ser to Ala cannot separate GSK3 from DYRK2 acting at the priming serine, and the catalytically dead form needs its own control in tissue with intact Gsk3b; an eIF2B activator only as a discriminating probe, with an engagement control; the null (the GSK3 route is not engaged) stated before the measurement. Lithium stays a labelled non-specific comparator. Source: a repository record (not included in this package).

   Scientific details of this row

   Origin of the proposal

   Carded thread: lithium.

   Available evidence

   WWOX-system with a specificity flag: one paper, one laboratory, the comparison drug reported better; the kinase output readout is a phosphorylation site in mouse plus a docking mechanism in cells. IT21 addition (P3b, proposal G02): the same provocation test in the null, heterozygous and wild-type littermates shows the lithium effect in all three genotypes, and a second documented action of lithium (inositol monophosphatase inhibition) has a documented pro-convulsant sign in rodents, which disfavours it as the mediator by analogy (the precedent is the lithium-pilocarpine model, not a measurement in the Wwox assay; Cheng 2020 also ran a pilocarpine model in Wwox mice, so the objection that the provocations differ is testable in the same system); abstract-level reading for the sign argument. [IT36] The ISR tool compound binds the eIF2B complex at the beta-delta interface and is a probe, not a candidate; the NB-2 discriminator by inhibitor class has partial precedents at abstract level and the two prime sources remain unread beyond abstracts; the ISRIB outcome grid fits several mechanisms and does not by itself show the GSK3-to-eIF2B chain.

   Development feasibility

   Selective tool compounds exist; the assay is the carded one.

   Novelty, within the verified scope

   Queried: this need's two observation rows; the lithium unit, its caveat rows and its counter-statements; the search layer over the four kinase-axis targets (588 associations, 526 families, none direction-coherent because the need direction is unresolved). The opposing action on the second need is recorded in this module's own need file.

   Reason for the move

   position unchanged from IT20 (rank 5): Position unchanged from IT20. The ten-row ceiling does not displace a row that carries a risk flag or human clinical reports in favour of a conditional probe (the rule stated in the curation block of the context), so this row stays at rank 5 and the two new IT21 conditional probes sit at rank 6, below the line. New IT21 input: the inositol-monophosphatase route to the lithium effect is DISFAVOURED_BY_SIGN_PRECEDENT (an analogy from the lithium-pilocarpine model, not a measurement in the Wwox assay; see what the generator rules out). That result is not what supports this row's concern that the effect is not specific to WWOX loss: the concern comes from the +/+ arm of Cheng 2020 Figure 7d, where lithium suppresses the seizure stage as in the null. Cheng 2020 also ran a pilocarpine model in Wwox mice, so the objection that the two provocations differ is testable in the same system; the inositol arm of that proposal is folded into this row's next test.

   Comparators

   3 functional-outcome, 7 mechanism-separating (compound names are not in the shareable edition)

   Conditional physiological hypotheses

   1 conditional hypotheses, with alternatives and what would strengthen or weaken them

   These are questions stated with their conditions, not acquired data: no compound is proposed here and no gene-to-gene link is inferred. The absence of an experiment is not demonstrated harm.

   • In WWOX-deficient neural cells with raised GSK3 activity, GSK3 inhibition lowers phospho-Ser535 (human Ser540) of eIF2B epsilon and raises exchange… (GSK3-H1; exploratory hypothesis)

     What is observed

     Not stated by the package as an observation.

     What is hypothesised

     In WWOX-deficient neural cells with raised GSK3 activity, GSK3 inhibition lowers phospho-Ser535 (human Ser540) of eIF2B epsilon and raises exchange activity and new-protein labelling beyond what the same compound does in the isogenic corrected cells. Exploratory: the package states no therapeutic direction.

     Alternatives that remain

     Alternatives the package states: (a) pS535 not raised at baseline: the chain breaks at L1 or L2; (b) pS535 falls but exchange and translation do not move (the CHO result, E06): site not sufficient; (c) effect also in corrected cells or unchanged by S535A: off-target (E07); (d) effect disappears when phospho-eIF2 alpha is lowered: the ISR route || Other rows of the package, not classified by it as rivals, complementary or compatible: H2: The relevant GSK3 isoform in WWOX-deficient neurons is beta, as in retinal ganglion cells, so a beta-directed knock-down reproduces… || H3: Replacing eIF2B epsilon by the Ser-to-Ala form (species numbering checked: human 540, mouse 536, rat 535) restores translation in… || H4: ISRIB (an eIF2B activator) is a discriminating probe, not an assumed rescue: if the defect is a GSK3-site lesion independent of… || H5: The GSK3 route is not engaged in WWOX-deficient neurons and ISR changes, if any, come from another input

     What would strengthen it

     a correction of translation and one functional read-out specific to the null (not seen in controls)

     What would weaken it

     (a) pS535 not raised at baseline: the chain breaks at L1 or L2; (b) pS535 falls but exchange and translation do not move (the CHO result, E06): site not sufficient; (c) effect also in corrected cells or unchanged by S535A: off-target (E07); (d) effect disappears when phospho-eIF2 alpha is lowered: the ISR route

10. 5 (tie) GABA-aminotransferase inactivators (see the vigabatrin card) thread already known in WWOX thread with a result already in a card#

    Question

    What does the published symptomatic response predict, and what does it not predict, about developmental outcome?

    Why it matters, what we know, next step

    Reason for the priority

    This is the thread with real human reports behind it and the one most likely to be read as a recommendation. The module already holds a series in which symptomatic response and developmental outcome did not move together, so the useful investigation is the separation itself, not another symptomatic observation.

    Solidity of the evidence (depth of reading)

    E3 - card reviewed with assertions; no WWOX-side result read beyond abstract

    Decisive uncertainty

    Whether the separation of symptomatic response and development seen in WWOX-DEE cases differs from the base rate of genetic and structural DEEs, in which a responder with poor development is the expected case (in unknown aetiology the two are coupled, so the base rate does not extend to DEE generally); whether any WWOX-DEE case shows a change on a continuous developmental measure after a symptomatic response; whether a sustained response, which no published child achieved, is followed by gain; and whether the separation holds for residual-function alleles. The earlier contradiction with the mechanism argument is not unexplained: other genetic DEE cohorts (STXBP1, Dravet, KCNQ2; partial full-text and abstract depth, not WWOX evidence) report a developmental component independent of seizure control. [IT36] caregiver-rated global improvement anchored at about 31 percent reduction (slight) and 49.6 percent (much) in Lennox-Gastaut; not WWOX evidence. [IT46] Correction to the base-rate wording: the cohorts support 'a majority outcome in infantile-spasm cohorts' (72 and 65 percent), not 'near-universal in genetic strata' (the genetic bound counts children who were not responders; one stratum has trisomy 21 in five of nine; two KCNQ2 cohorts show 67 and 80 percent); in the largest cohort 145 of 175 responders also received prednisolone, and in another 39 of 61 responders relapsed. Separately, a GABA-raising lever can aggravate absence-type seizures, so any effect is seizure-type dependent before it is lesion-specific. Source: a repository record (not included in this package) (R6, R7). [IT47 SCI47] Unit-of-analysis note for the Wwox slice paper: the change in inhibitory-current frequency is shown only by a distribution test on pooled events (no per-cell or per-animal test); the pipette of the excitatory-current recordings holds about 12 mM chloride (E_Cl about -63 mV), so their isolation at -70 mV is not chloride-free; burst frequency differs between the Results (median interval 26.2 s) and the supplementary table (0.84 per minute); animal counts do not sum (4+4 against 7). Source: a repository record (not included in this package).

    Next test

    EXECUTED 2026-11-05: the published same-genotype cases with both a symptomatic and a developmental endpoint were assembled (6 children with both endpoints; a discordance rate is not computable at n below 10, so counts are given instead). Vigabatrin fired the 5-or-more branch (counts with sources); hormonal therapy, diet and cannabidiol fired the fewer-than-5 branch (not computable from the public record). Next: read these counts against the stratum base rates in a repository record (not included in this package) section 1.3; only a WWOX-DEE responder rate with a favourable or improving developmental course clearly above the genetic-DEE base rate would be a WWOX finding. Denominator request, without contact: for each WWOX-DEE series that reports an intervention, the number exposed, the response definition, and a dated developmental instrument before and after a sustained response; where a source gives one, compare within-child continuous scores, since a cut at 70 cannot discriminate. The five-patient series and Choi 2026 come first, and the seven-child series (PMID 42721537) is read when an open route exists. No new exposure is proposed. [IT36] the 12-child series (2 seizure-free, 9 continuing, 1 death) is unread for treatment; its two seizure-free children are the first candidates for a sustained response. [IT46] For the mechanism question (S7): measure the GABA reversal potential by genotype at P13 to P17 against rest, threshold and conductance before any compound arm; then a GABA-supply arm against an unrelated arm of equal measured effect (compound view V20). [IT47 layers] For any antiseizure drug named on this thread three layers are kept apart: clinical evidence (what is recorded in children with WWOX-related disease; here single reports without an instrument), preclinical rationale (what an experiment in a WWOX model shows) and untested hypothesis (a mechanism proposed from outside WWOX). An antiseizure effect does not by itself support a preference or a better development; a drug already in clinical use may deserve a comparison only if a WWOX-specific rationale emerges. The order of the investigations is an investment in research, not an expected benefit for a child.

    Compounds or classes involved (research order)

    • still unresolved Raising GABA availability (reuptake or transaminase block): the sign of the action at P13 to P17 is unmeasured and depends on driving force and shunting
    • still unresolved Positive modulation of the GABA-A receptor: spontaneous inhibition is measured as reduced in the recorded cells, its cause and its sign are not

    Scientific details of this row

    Origin of the proposal

    Carded clinical thread: GABA-transaminase inactivators.

    Available evidence

    Clinical reports only. Children counted once across 14 reports: 36 to 38 exposed to at least one of vigabatrin, hormonal therapy, ketogenic diet or cannabidiol (partial full-text depth in 11 reports, one source at abstract level). Both a symptomatic and a developmental outcome are recorded for 6. In 4 of them, a partial or transient improvement came with no developmental progress recorded; in 2, neither moved. No sustained, instrumented response is recorded, and no developmental gain follows any of the four classes. Vigabatrin: 5 children with both endpoints in 4 reports (4 discordant, 1 concordant non-response); these are not independent of the other classes, since three of the five also supply them and all five were on co-therapy. Children with both endpoints, by class, of those exposed: hormonal 3 of 11; ketogenic diet 2 of 12; cannabidiol 1 of 7. Counts only, no percentage (n below 10). Developmental statements are clinical narratives without a scale. A brain-imaging safety signal is recorded beside them in two children (abstract level, a report selected for the finding), so it is not a frequency. No WWOX child is recorded with a developmental gain after a symptomatic improvement. Four children in three reports (two of them in one report) had a partial, transient or clinically-defined improvement with development not progressing; two had neither. That is what the baseline in other genetic and structural early-onset encephalopathies predicts (a responder with poor development is the majority outcome in infantile spasms and near-universal in genetic strata), so it is not by itself a WWOX-specific feature. The comparison holds only as a binary: the baseline uses standardised scores with a cut, every WWOX developmental statement is a clinician narrative, and no WWOX child had a sustained, instrumented response. Non-WWOX DEE baseline (batch_sci_20261105 a2_baseline; not WWOX evidence): in a prospective IESS cohort read at partial full-text depth, 72 percent of responders (126 of 175, computed from its Table 1) were below the developmental cut at 3 years, and no genetic-aetiology child reached a favourable outcome (0 of 49, a count over all genetic-aetiology children, not a responder rate). In two randomised IESS trials (abstract level), arms that raised early spasm cessation by 15 to 19 points showed no demonstrated developmental difference, with about one chance in four of detecting the difference that full mediation would predict. In TSC, two randomised trials of pre-onset prevention changed seizure endpoints without changing development to 36 months; they test prevention before onset, not response after onset. The person-level association (ICISS, abstract level: 15.9 VABS points higher with early cessation) is carried by unknown aetiology and a normal first neurological examination. [IT36] the 2024 vigabatrin case: developmental statements uninstrumented, stall from spasm onset before drug; seizure-free at home while video-EEG still showed events. [IT39] The transfer from KCNQ2 and TSC is analogy; genotype stratification is established for mortality only; lithium is not the only existing-drug node (ethosuximide was also tested).

    Development feasibility

    A record-level analysis on material already in the module; no experiment.

    Novelty, within the verified scope

    Queried: this need's three observation rows; the inactivator unit, its caveats and the safety flag; the chloride-gradient and inhibitory-hypofunction needs for the mechanism argument that contradicts the reports. Extended on 2026-11-05/06: the non-WWOX DEE baseline queried (UKISS, ICISS, the National Infantile Spasms Consortium records with no developmental outcome endpoint, lead-time cohorts, EPISTOP and PREVeNT, STXBP1, Dravet and KCNQ2 cohorts, and an early-infantile DEE cohort with burst suppression; a repository record (not included in this package)); 14 WWOX-DEE reports queried for exposure and both endpoints (a1_wwox_reports). Bounded queries, not a statement that nothing else exists.

    Reason for the move

    position unchanged from IT20 (rank 5): Position unchanged from IT20. The ten-row ceiling does not displace a row with real human clinical reports in favour of a conditional probe (curation rule), so this row stays at rank 5 and the two new IT21 conditional probes sit at rank 6, below the line. This row is a record-level analysis that needs no experiment budget and can proceed alongside.

    Comparators

    6 functional-outcome, 0 mechanism-separating (compound names are not in the shareable edition)

Below the ten-row limit (in the full table, with their question and their test): INV-12; INV-13; INV-15; INV-11; INV-14; INV-16; INV-17; INV-18; INV-19

Contribution of WWOX to the measurement. Retrospective measurement (IT21 P5, re-read in IT22): replacing the WWOX starting genes with randomly chosen genes of the same network size, the generator produces almost as many proposals (none of the 5 sets exceeds the 95th percentile of the closest comparisons). The conclusion holds only for the properties measured (number of proposals and membership of random sets): it is not the general conclusion that WWOX adds no value, because network interactions, cell-type-specific contrasts and premises about the role were not measured. It says nothing about prospective benefit, nor that the compounds listed are inactive. It reopens with a datum measured in the WWOX system that distinguishes the proposals.

A novelty holds only within the declared search scope: absence from a card or from a search does not mean “never studied”.