{"schema_version":1,"assessment_id":"eoa_inverse_innovation_exp03_opportunity320_20260801","source_experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"invariant_mode_decomposition_design__pharmacology_toxicology","archetype_slug":"invariant_mode_decomposition_design","domain_slug":"pharmacology_toxicology","title":"Harm-weighted transition-mode detection for repeat-dose toxicity","opportunity_summary":"Apply a bounded longitudinal transition model and eigendecomposition to completed repeat-dose toxicology data to test whether persistent or growing joint measurement trajectories reveal consequential toxicity earlier or more accurately than coordinate-by-coordinate review and a direct multivariate predictor. Modal outputs remain hypotheses, require residual, conditioning, spectral-gap, identity, and drift checks, and cannot independently change exposure or sampling.","adopter_authorizer":"The toxicology study director can authorize a retrospective shadow analysis. Any prospective exposure or sampling change requires the protocol owner, veterinarian, animal-welfare body, and organizational safety governance.","scores":{"meaningful_impact":{"score":4,"rationale":"Earlier recognition of consequential toxicity could improve safety-margin estimation, monitoring, and dose selection for animals and downstream participants. Impact is potentially substantial, but the sealed candidate provides no evidence about how often coupled subthreshold trajectories occur or how much harm ordinary review misses."},"stakeholder_pull":{"score":3,"rationale":"Study directors, sponsors, welfare authorities, and downstream safety decision-makers have a plausible interest in earlier detection without extra false alarms. Actual demand, workflow pain, prevalence, and willingness to adopt are unsupported and require external inquiry."},"incremental_advantage":{"score":3,"rationale":"The proposal could add interpretable persistence, growth, modal-stability, and spectral-failure diagnostics beyond univariate review and direct multivariate prediction. No sealed evidence shows earlier detection, fewer false alerts, or better decisions than either comparator."},"distinctiveness_plausibility":{"score":2,"rationale":"The composition of invariant transition modes, harm-weighted sensitivity, and explicit spectral validity gates is structurally specific, but prior art is unsearched and distinctiveness from longitudinal multivariate, state-space, PK/PD, and toxicology safety methods is unverified."},"technical_implementability":{"score":3,"rationale":"A retrospective shadow analysis on an existing longitudinal study is technically bounded and avoids live decision changes. Feasibility is constrained by sample size, missingness, scaling, correlated error, nonlinear thresholds, non-normal transients, sparse observations, changing metabolism, and potentially unstable or poorly separated modes."},"adoption_authority_feasibility":{"score":4,"rationale":"The candidate names a study director able to authorize decision-neutral shadow analysis and separately names the authorities required for prospective protocol changes. Broader organizational or regulatory acceptance remains unverified."},"evidence_readiness":{"score":4,"rationale":"The packet specifies a preregistered completed-study test with designated fitting data, blinded held-out observations, fixed outcomes and thresholds, two comparators, and explicit performance and validity checks. Availability and suitability of a completed dataset are assumed rather than demonstrated."},"safety_net_benefit":{"score":4,"rationale":"The proposal preserves ordinary review, prohibits model-only exposure changes, requires confirmatory evidence, and specifies halt and rollback conditions. These controls materially contain risk, although false reassurance, modal compression, and mechanism overinterpretation remain possible."},"scalability":{"score":2,"rationale":"Software analysis could be repeated, but each compound, exposure regimen, measurement panel, population, and clinical setting may alter dynamics and requires revalidation. Data harmonization, spectral stability, governance, and rare-signal preservation limit straightforward scaling."}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"50K_TO_250K","scope":"One preregistered retrospective shadow analysis on a completed repeat-dose study, including data curation, outcome and threshold locking, model and comparator implementation, blinded held-out evaluation, robustness checks, toxicology review, and reporting.","confidence":"MODERATE","assumptions":["Usable longitudinal measurements and adjudicated outcomes already exist.","No new animals, samples, assays, or prospective protocol changes are required.","The work uses a small multidisciplinary team and existing general-purpose computing.","Data-access and contracting burdens are limited to one cooperating organization."]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"Build a governed, validated shadow-use workflow across several studies, including data pipelines, auditability, model validation, prespecified failure gates, user review procedures, documentation, training, and integration with ordinary toxicology review.","confidence":"LOW","assumptions":["Deployment remains advisory and does not autonomously alter protocols.","Several representative historical studies are accessible for validation.","No new laboratory platform or major instrumentation is purchased.","Quality, compliance, and integration requirements are substantial but remain within one organization."]},"operational_launch":{"band_2026_usd":"1M_TO_5M","scope":"Launch prospective governed use across multiple repeat-dose studies, including protocol and welfare review, confirmatory assay planning where approved, prospective comparator evaluation, monitoring, validation by context, and operational support.","confidence":"LOW","assumptions":["Prospective use requires partner coordination and formal governance.","Any added sampling or procedures receive separate approval and are included in the resource range.","Multiple compounds or study contexts require distinct validation work.","The launch evaluates decision support rather than granting autonomous control authority."]},"annual_recurring":{"band_2026_usd":"250K_TO_1M","scope":"Maintain an organizational decision-support service with data operations, study-specific recalibration and revalidation, drift and failure monitoring, quality review, user support, documentation, and periodic comparator audits.","confidence":"LOW","assumptions":["Use spans multiple studies per year within one organization.","Existing laboratory and toxicology operations continue separately.","Material model changes or expansion to clinical populations would require additional project funding.","Recurring costs depend strongly on study volume and data standardization."]}},"research_burden":"HIGH","earliest_credible_horizon":"3_TO_12_MONTHS","pipeline_gates":{"recognizable_externally_supportable_problem":{"status":"UNCERTAIN","reason":"The candidate clearly states a falsifiable problem—coupled subthreshold measurements may conceal toxicity—but supplies no external evidence that it occurs with decision-relevant frequency or defeats ordinary review."},"identifiable_adopter_or_authorizer":{"status":"YES","reason":"The toxicology study director is identified as the shadow-analysis authorizer, with protocol owners, veterinarians, animal-welfare bodies, and organizational safety governance identified for prospective changes."},"distinct_testable_incremental_claim":{"status":"YES","reason":"The proposal can be tested for earlier or more accurate held-out toxicity detection and acceptable false alerts versus coordinate-by-coordinate review and a direct multivariate predictor, while separately checking modal stability and residual validity."},"bounded_next_evidence_step":{"status":"YES","reason":"A preregistered, blinded, decision-neutral analysis of one completed study is explicitly bounded by fixed outcomes, designated fitting data, comparators, validity checks, and halt criteria."},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"The first step changes no decisions or animal procedures, preserves ordinary review, and includes explicit prohibitions, halt conditions, rollback, and named authorities for any later prospective action."},"implementation_cost_scope_and_range":{"status":"YES","reason":"A one-study retrospective analysis and later governed deployment can be scoped into broad resource bands, although actual data condition, compliance requirements, study volume, and prospective procedures create substantial range uncertainty."}},"blocking_evidence":["External evidence that coupled subthreshold dynamics are a decision-relevant problem and are missed or detected late by ordinary review.","Blinded held-out evidence that the method improves detection timing or accuracy without exceeding baseline false-alert tolerance.","Evidence that modes have acceptable conditioning, spectral separation, stable identity, drift behavior, and harm-weighted residuals across replicates or windows.","A bounded prior-art comparison establishing whether transition-mode stability, harm weighting, spectral validity gates, and governed toxicology use are distinct from existing methods.","Evidence that performance persists after accounting for exposure, time, missingness, scaling choices, measurement error, nonlinear behavior, and non-normal transient amplification.","Before action-oriented adoption, prospective evidence that mode-ranked assays or governed dose, schedule, or monitoring hypotheses improve outcomes over the nearest rival."],"next_evidence_step":"Run the specified preregistered shadow analysis on one completed repeat-dose study: fit only on designated animals and early time points, blind held-out outcomes, and compare detection timing, consequential-event accuracy, and false alerts against univariate review and a supervised multivariate or conventional PK/PD rival. Falsify advancement if ordinary review is at least as early and accurate, no repeatable coupled signal remains after exposure, time, and measurement-error adjustment, or conditioning, spectral separation, modal identity, drift, or harm-weighted residuals breach preregistered limits.","research_questions":["How often do consequential repeat-dose toxicity trajectories involve coupled measurements that remain below individual alert thresholds?","Does the modal method detect held-out consequential toxicity earlier or more accurately than ordinary review and the nearest multivariate rival at an acceptable false-alert rate?","Are candidate modes reproducible across animals, replicate datasets, resampling, preprocessing choices, and local time windows?","What minimum sampling density, sample size, spectral separation, and conditioning are required for usable modes?","How sensitive are results to missingness, scaling, correlated measurement error, exposure adjustment, and endpoint selection?","Do nonlinear thresholds, irreversible lesions, changing metabolism, or non-normal transient amplification invalidate local eigenvalue-based interpretation?","Does modal compression suppress rare but consequential signals or create false reassurance?","Can toxicologists use mode-ranked confirmatory assays without treating mathematical loadings as biological mechanisms?","Are the proposed composition and validity gates distinct from existing longitudinal multivariate, state-space, PK/PD, and toxicology safety approaches?","In a later governed prospective test, do mode-ranked hypotheses improve detection, prediction, or control over the nearest rival?","What study director, welfare, quality, sponsor, and regulatory approvals are required at each level of use?","What data preparation, integration, compliance, and monitoring burden is required for repeatable operation?"] ,"recommendation":"PRIOR_ART_RESEARCH","uncertainty_constraints":["This is a closed-book assessment with no external validation of prevalence, stakeholder demand, adoption appetite, market size, realized impact, or costs.","The candidate is a hypothesis; no retrospective or prospective performance evidence is supplied.","Prior-art status is explicitly unsearched, so structural specificity cannot support a novelty conclusion.","Mathematical modes are not established biological mechanisms or independently controllable processes.","Resource bands are broad scenario estimates conditioned on existing data and organizational infrastructure, not exact costs.","Generalization across compounds, regimens, measurement panels, species, populations, or clinical settings cannot be assumed.","The earliest horizon applies only to retrospective shadow evidence, not operational or action-oriented adoption."],"closed_book_prior_art_boundary":"Prior art is unsearched and unverified in this closed-book assessment. No claim is made about novelty, freedom to operate, prevalence of similar methods, or distinctiveness from existing longitudinal multivariate toxicology, state-space safety, supervised prediction, or PK/PD approaches; those questions require external research."}