Classification, burden, and treatment framework

This volume orders systemic genetic and rare diseases by a deterministic burden proxy built from five graded axes -- onset, progression, severity, mortality, disability -- then offset by the established therapy's efficacy. Clinical facts are observed, cited inputs; the classification and the burden order are the reproducible contribution. The order is a provisional prioritisation device, not a registry-locked ranking.

Each disease is placed on three orthogonal axes (inheritance, molecular mechanism, organ system) and scored on a fixed a-priori burden scale whose composite is burden_score = raw_burden · (1 − e). Every value carries a grade: [V] law, [L] registry/measured, [H] cited-text inference, [O] open with a named obstacle.

Observed inputs vs the reproducible layer

Clinical and epidemiologic facts here are observed inputs from NCBI (MedGen, OMIM, Gene, ClinVar) and HPO: respected and cited, never asserted as “reproduced”. The volume’s reproducible contribution is the analysis layer — the inheritance / mechanism classification, the deterministic burden index, the residual-treatment offset, and the normal-development engine annotation.

Reproducibility is bit-for-bit: the burden order is regenerated by an in-package pipeline (r3_burden_index.py → r4_burden_residual.py) and a gate re-runs the builders and asserts an identical digest (2× sha256). Every number on every disease page is read from that pipeline’s output, not typed into prose.

Three classification axes

Each entity is placed on three orthogonal axes so that “how it is inherited” is never confused with “what the variant does to the protein”.

The sibling neuro/mind whitepapers own the brain, nerve, heart, and affect. A multi-system disease is classified by its primary in-scope system; its excluded-organ features are named and cross-referenced, not re-described (SCOPE.md).

The deterministic burden index

Subjective suffering is individual and not directly measurable, so this volume does not rank suffering. It defines a transparent burden proxy from observable, published clinical attributes — a prioritisation device for coverage and triage, explicitly not a value judgement about any patient.

Five axes are each binned to [0,1] by fixed a-priori clinical meaning (tier values are not chosen to produce any ordering):

The composite is a renormalised mean over the axes actually scored (a missing axis is excluded, never imputed): raw_burden = mean(scored axes), and burden_score = raw_burden · R_treat. Equal weights of 0.20 are declared and recorded.

Honest limitation, front and centre. Onset is registry-grade [L] for most of the cohort, taken from the Orphanet AverageAgeOfOnset register (earliest typical category, entity-anchored per ORPHAcode, Exact OMIM↔ORPHA only) cross-checked with HPO (R6). Disability is now registry-grade [L] for the diseases whose one dominant untreated sequela maps to a named GBD 2013 health state, the published disability weight binned by declared cut-points (R7); two mortality axes are independently corroborated from PMC survival literature. Severity now lifts to registry [L] for the one disease whose dominant sequela carries a cited annotation in the HPO clinical-modifier Severity subtree (HP:0012824, R8); for the rest of the cohort the open HPO severity annotations are feature-level (a category error if read as a disease tier) and the OMIM clinical synopsis is API-key-gated with the key unobtainable for an individual researcher, so severity stays largely [H]/[O]. Progression lifts to registry [L] where a cited PMC open-access source states a disease-level magnitude for the dominant untreated sequela and the frozen R3 tier function derives the tier from that verbatim sentence (R9, non-spectrum, non-comparative dominant-sequela join); for the rest progression stays [H] inferences from the cited definition. The whole order is therefore a provisional [H] prioritisation device, not a registry-locked ranking. 6 diseases (Achondrogenesis type II, Duchenne muscular dystrophy, Marfan syndrome, Niemann-Pick disease, type A, Phenylketonuria, Tyrosinemia type II) now have every one of their scored axes at registry [L] (order_locked, 6 of 35): their burden value is registry-grade, though the cohort order overall stays provisional until severity and progression lift further via published functional & survival literature — not guessed.

Rankability

A mean over fewer than a majority of the five axes is not a meaningful aggregate, so one fixed declared cut applies: rankable = (axes_scored ≥ 3). Of the cohort, 23 diseases are placed in the contiguous order on this page; the rest are reported as “not placed — insufficient axis coverage”. This is a declared coverage cut, not a tuned parameter.

Treatment evidence and the residual offset

The clinically relevant quantity is residual burden — what remains after the best established therapy. Each disease’s established disease-directed therapy is surveyed, its mechanism stated only where mechanistically established (C-D3), and an evidence status mapped to a fixed efficacy offset e.

Under the grading rule the treatment modality itself is never assigned a fabricated cure: its mechanism is stated only where mechanistically established (C-D3). The natural-history effect of each established therapy has been accession-dated to the GeneReviews Management section [L] for 33 of 35 diseases (the R5 pass; initial-posting and last-revision dates recorded per page); the two without a current GeneReviews chapter stay [H]. Cross-dating the same effect against the FDA label / OMIM clinical management is the deferred extension. The residual order remains provisional because the burden axes — not the treatments — are still a mix of [L] and [H].

The carried normal-development engine

Many developmental diseases are perturbations of a morphogenetic programme. This volume carries, read-only, a DB-grounded morphogenesis engine that forward-simulates normal development from measured, cited biophysics and grades every quantity honestly.

Grade legend

The rankable burden order

The placed cohort, by residual burden (each disease is its own page):

All 23 residual figures are [H] provisional. Diseases with 1–2 scored axes are not placed and are authored separately in later sections.

Forward plan and completion criteria

This volume is built in versioned, gated passes, and the forward plan is carried inside the whitepaper itself — in every version — until an explicit completion declaration replaces it. Each pass either lifts a burden axis to registry grade through a cited, gate-enforced rule, or records, per disease, the precise obstacle that prevents it. No value is ever guessed to advance the plan.

Landed so far. Treatments accession-dated [L] (33 of 35); onset [L] (28 of 35, Orphanet, R6); disability [L] (13 of 35, GBD 2013, R7); mortality [L] (8 of 35, PMC survival literature, R7); severity [L] (12 of 35, HPO Severity subtree HP:0012824 R8 + curated PMC open-access R11+R12+R14, frozen-R3-derived tier); progression [L] (12 of 35, curated PMC open-access literature with a frozen-R3-derived tier, R9+R10+R13+R14). order_locked 6 of 35 (Achondrogenesis type II, Duchenne muscular dystrophy, Marfan syndrome, Niemann-Pick disease, type A, Phenylketonuria, Tyrosinemia type II). A read-only unmet-need / treatment-gap surface (residual = raw_burden · (1−e)) re-presents the same registry as a graded research-prioritisation signal.

Next task (R16). R9–R15 took the curated-PMC-OA frontier through the clearly defensible disease-level joins reached so far: R9’s progression lifts (NPD-A locking), R10’s two grade-only progression corroborations (Fabry and Marfan, [H]→[L] value unchanged, neither locking), R11’s single severity lift (Marfan syndrome [H]→[L], completing the 4th order-lock), and R12’s two severity lifts — Duchenne muscular dystrophy severity [H]→[L] (0.50→0.75, tier derived by the frozen R3 function via ‘severe’ from the disease-defining sentence) which, because Duchenne’s onset, progression, mortality and disability were already [L], completes its order-lock (the 5th), plus Tyrosinemia type I severity [H]→[L] (value 0.75 unchanged, a grade-only lift that does not lock, two [H] axes remaining). R12 also recorded two declines, each a cited open-access sentence found and verbatim-pinned but not a clean disease-level magnitude: Hurler syndrome severity (comparative — ‘the severe phenotype’ of MPS I in contrast to the attenuated forms) and the beta-thalassemia umbrella severity (umbrella spectrum — silent/minor → intermedia → major, tripping the R3 spectrum override). A frontier observation is recorded for hemophilia A: its only retrievable open-access ‘severe’ framing is ‘severe hemophilia A (FVIII<1%)’, a factor-activity sub-phenotype of an entity spanning mild/moderate/severe, so no disease-level severity tier is isolable and severity is left [H], not forced. R13 then added three lifts — Maple syrup urine disease severity [H]→[L] (via ‘life-threatening’) plus Cystic fibrosis and Pompe progression grade-only — and R14 added three more — Classic homocystinuria and Fabry disease severity 0.50→0.75 (via ‘life-threatening’ / ‘severe’) plus Becker muscular dystrophy progression grade-only (via ‘progressive’, advancing the R13 BMD frontier on a different axis from a clean BMD-specific sentence) — none completing a new order-lock (each lifted disease retains another [H] axis, so the lock set stayed 5). R14 also declined Polycystic kidney disease 2 severity (comparative, PKD1-vs-PKD2) and Osteogenesis imperfecta progression (spectrum, moderate-to-mild across a 17-patient cohort). R15 then added three severity lifts — Acute intermittent porphyria [H]→[L] (0.50→0.75, via ‘severe’ from the disease-defining neurovisceral-attack sentence; stays sub-rankable at 2 scored axes, so no lock), Achondroplasia [O]→[L] (0.75, via ‘severe disproportionate short stature’; becomes rankable but retains mortality [H], so no lock), and Phenylketonuria [O]→[L] (0.75, via ‘severe intellectual disability and irreversible brain damage’ in the untreated-classical-PKU sentence) which, because PKU’s onset and disability were already [L], completes the 6th order-lock. That lock is disclosed as an honest consequence, not lock-fishing: PKU’s severity and disability axes are orthogonal by construction (disease-severity magnitude vs the GBD functional-dependence weight), the sources differ, and the cited sentence passes every inclusion criterion independently — so the lock set moves 5→6 with the delta exactly +Phenylketonuria (the gate asserts this). R15 also declined 21-hydroxylase-deficient congenital adrenal hyperplasia severity (form-specific — ‘life-threatening salt-wasting’ attaches to the severest classic form, not the entity that also spans the mild non-classic form) and Hemochromatosis type 1 severity (comparative — ‘severe clinical outcomes’ framed only as a high-vs-low-TSAT between-stratum contrast). With 17 placed diseases still carrying an [H] axis — overwhelmingly severity — R16 continues the same rule on the placed cohort’s remaining [H]/[O] severity and progression axes (and any further mortality axis): a disease-level magnitude for the dominant untreated sequela, with the tier derived by the frozen R3 tier function over the verbatim cited sentence — non-spectrum, non-comparative, non-treated-cohort, non-sub-phenotype. The cut-points are never widened; every lift is gated and every decline is recorded with its reason; the pass ships as a new cumulative stage that leaves all prior artifacts byte-identical. Mortality lifts only where a quantitative disease-typical survival figure exists (free text does not fill an open mortality axis). The OMIM clinical-synopsis path stays removed (its key is unobtainable for an individual researcher), not deferred.

Completion declaration (the criterion). This volume will be declared complete when, for every placed disease, either (a) all scored axes are registry grade [L]/[V] (the disease is order_locked and its burden value is registry-locked), or (b) every remaining open axis carries a permanent, named obstacle that is irreducible from open data (e.g. a magnitude that exists only behind an unobtainable licensed source). At that point the burden order is either fully registry-locked or provably as locked as open data allows, the residual [H] openness is itemised in IRREPRODUCIBILITY_LEDGER.md with each obstacle, and a dated completion declaration replaces this forward plan. Until then, this section is part of the whitepaper in every version.