Curing Lupus Atlas

Operational research roadmap

Turn hypotheses into studies that can fail honestly.

Three protocol packages define the next testable programs, the gates that keep them interpretable, and the work that can proceed before any human-subject activity is authorized.

Inspect the programsSee the sequence

Program architecture

Different questions. Shared safeguards.

The programs share governance and infrastructure, but their results cannot be pooled into a single causal claim.

Shared foundationGovernance · provenance · assay contracts
01Predict transition02Explain relapse03Match mechanisms
Conditional handoffMechanism-isolating experimentOnly after the relevant gates pass

Protocol packages

What can move—and what must wait

Preparatory work is real progress. It does not imply permission to enroll, contact participants, transfer protected data, or change care.

01

protocols/pre-disease-validation/

Pre-disease transition validation

Can a locked, transportable model estimate SLE-specific transition risk accurately, fairly, and usefully enough to justify a later mechanism-matched prevention trial?

Current gateGate 0 · Protocol readiness

No model may advance until predictor timing, competing outcomes, leakage, calibration, participant governance, and independent external validation are operational.

Can start now

  • Audit cohort provenance, screening logs, predictor timestamps, outcome definitions, and participant overlap.
  • Validate the common data dictionary, synthetic fixtures, missingness rules, and reproducible analysis tables.
  • Inventory low-cost clinical predictors before adding molecular layers.
  • Prepare risk-communication and false-positive-harm review packets for accountable experts and patient partners.

Cannot start yet

  • Recruit, disclose individual risk, or label ANA positivity as a high-risk state.
  • Test a preventive intervention or use a biomarker to direct care.
  • Call case-control discrimination an absolute-risk model.
Decisive output

An independently replicated, calibrated absolute-risk model with net benefit beyond low-cost clinical predictors—or a documented stop decision showing that prevention-trial recruitment is not justified.

A useful failure: It prevents an inaccurate model from creating false-positive surveillance, anxiety, inequity, or an unsafe prevention trial.
Inspect roles & ownership

Fable / non-domain operations

  • Cohort and screening-log inventory
  • Timestamp and leakage checks
  • Schema and fixture validation
  • Decision-ledger maintenance

Accountable domain owners

  • Rheumatology
  • Epidemiology
  • Prediction methods
  • Ethics
  • Patient/community
02

protocols/reset-relapse-observatory/

Reset–relapse mechanistic observatory

Which measured tissue, plasma-lineage, clonotype, and reconstitution states predict durable drug-free remission or precede relapse after one fixed immune-reset package?

Current gateGate 0 · Human and regulatory authorization

No enrollment or specimen collection is authorized. A sponsor, investigators, ethics boards, safety monitors, patient partners, and regulators must approve an implementation protocol.

Can start now

  • Map existing cohorts and overlapping publications into the common study-family registry.
  • Inventory already-consented blood, tissue, marrow, repertoire, outcome, and safety data.
  • Test data dictionaries, visit-window logic, assay metadata completeness, and synthetic analysis pipelines.
  • Prepare cross-center outcome, specimen, and adjudication harmonization packets.

Cannot start yet

  • Enroll participants, collect extra tissue, or change a host trial’s treatment, rescue, or withdrawal plan.
  • Delay rescue care for relapse sampling or return unvalidated biomarker results.
  • Interpret blood absence as tissue eradication, naive phenotype as tolerance, or remission as cure.
Decisive output

A prospectively validated predictor or mediator candidate that precedes relapse and survives overlap, exposure, missingness, tissue-indication, and protective-immunity analyses—ready, if warranted, for a separate mechanism-isolating trial.

A useful failure: It demotes a reservoir or reconstitution theory before an unsafe or uninterpretable perturbation study is attempted.
Inspect roles & ownership

Fable / non-domain operations

  • Study-family and overlap registry
  • Visit and specimen completeness checks
  • Versioned outcome derivations
  • Audit-ready decision packets

Accountable domain owners

  • Rheumatology
  • Immunology
  • Nephrology
  • Repertoire science
  • Safety/ethics
03

protocols/frozen-selector-reanalysis/

Frozen-selector randomized-trial reanalysis

Does a frozen eight-endotype whole-blood selector reproducibly modify randomized treatment effect beyond ordinary clinical, serologic, continuous IFN, and BAFF/B-cell-state measures?

Current gateGate 0 · Executable artifact lock

Exact model objects, class maps, gene sets, transformations, continuous control formulas, reference distributions, licenses, fixtures, and cryptographic hashes are not yet available.

Can start now

  • Audit trial provenance, participant overlap, specimen timing, consent, endpoint maps, and data-access feasibility.
  • Build synthetic-data tests, signed analysis containers, mock tables, and decision ledgers.
  • Request and independently verify executable selector and control-measure artifacts without viewing outcomes.
  • Inventory nonoverlapping randomized trial samples and designate only provisional roles.

Cannot start yet

  • Reconstruct, retrain, relabel, merge, split, or recalibrate the selector on an outcome-bearing trial.
  • Unlock treatment or outcomes, declare a trial independently eligible, or run a confirmatory interaction.
  • Use the selector to allocate care or describe IFN/BAFF as stand-alone cure targets.
Decisive output

A frozen treatment-by-marker interaction that passes multiplicity, clinical-magnitude, calibration, added-value, equity, and independent replication gates—or falsification of the selector’s treatment-predictive use.

A useful failure: It distinguishes reproducible molecular grouping from actual treatment-effect prediction and prevents post hoc subgroup stories from guiding care.
Inspect roles & ownership

Fable / non-domain operations

  • Artifact manifest and hash ledger
  • Trial provenance and overlap audit
  • Synthetic test harness and table checks
  • Deviation and custody-log completeness

Accountable domain owners

  • Trial methods
  • Biostatistics
  • Laboratory science
  • Rheumatology
  • Privacy/ethics

Shared sequencing

Evidence unlocks the next phase—not the calendar

Indicative windows organize preparation. Event accrual, safety, validation, and gate decisions determine actual progression.

  1. 01
    Phase 0

    Make the work executable

    Planning window: 0–3 months after accountable leads exist

    Establish governance, provenance, exact assay/model artifacts, common definitions, data feasibility, simulations, and independent gate review.

    What it unlocksA public feasibility report and permission to begin bounded retrospective work—not human-subject research.
  2. 02
    Phase 1

    Use existing data without contaminating validation

    Indicative window: 3–12 months after Phase 0 passes

    Bridge assays blindly, freeze analyses, test existing cohort feasibility, and publish exactly what current specimens and records can and cannot answer.

    What it unlocksExternally testable artifacts, trial/cohort roles, and evidence-based decisions about prospective investment.
  3. 03
    Phase 2

    Prospective observational validation

    Duration follows event accrual—not a calendar promise

    Validate transition risk and reset–relapse states prospectively with blinded adjudication, standardized sampling, safety, and population-performance monitoring.

    What it unlocksTransportable prediction and candidate mediation; still not proof that changing the marker changes outcome.
  4. 04
    Phase 3

    Run mechanism-isolating experiments

    Only after the relevant protocol gates pass

    Test prevention or remission mechanisms with randomized clinical outcomes, direct target engagement, protective-function plans, and explicit falsification.

    What it unlocksA causal estimate capable of raising—or rejecting—cure relevance in a defined population.

Before all three programs

Build the foundation once

Shared infrastructure prevents the same participants, assays, endpoints, and uncertainties from being counted differently in each workstream.

  1. 01

    Accountable scientific, statistical, laboratory, clinical, ethics, safety, privacy, operations, and patient/community owners

  2. 02

    Study-family and privacy-preserving participant-overlap registry

  3. 03

    Versioned assay/model contracts, fixtures, hashes, and independent execution rights

  4. 04

    Common outcome, safety, missingness, specimen, and late-follow-up definitions

  5. 05

    Publication independence, including null, harmful, and inequitable results

Operational participation

Fable can move the machinery—not decide the biology

Non-domain contributors can make the work reproducible and reviewable. Scientific, clinical, safety, ethics, and authorization decisions remain with qualified accountable humans.

In scope now

Operations, evidence hygiene & QA

  • Track provenance, cohort families, participant overlap, metadata, dependencies, and deadlines.
  • Validate schemas, synthetic fixtures, tables, calibration displays, audit trails, and version changes.
  • Prepare reviewer packets and plain-language summaries using already approved uncertainty statements.
  • Monitor registries, publications, corrections, and required-field completeness.
Escalate to accountable humans

Domain judgment & authorization

  • Assign pathogenicity, biological causality, clinical eligibility, or target priority.
  • Judge intervention or adverse-event safety, define rescue, or approve risk disclosure.
  • Authorize enrollment, specimen access, participant contact, data transfer, or human-subjects research.
  • Turn a passing software check into scientific, clinical, ethics, or regulatory approval.

Human-subject boundary: This roadmap is not ethics approval, regulatory authorization, participant consent, funding, a data-use agreement, or permission to contact anyone. Passing a software or completeness check authorizes no study activity.

Review the target portfolioRead the editorial methods