Curing Lupus Atlas

Hypothesis lab

Turn connections into questions we can test

A disciplined workspace for synthesis—not a protocol, prediction, or recommendation for personal treatment.

Candidate connections

Hypotheses with a way to be wrong

Each record separates rationale from evidence, names a predicted result, and specifies what would weaken the idea.

Site hypothesisVerified 2026-07-10

Does the composition and autoreactivity of the reconstituting B-cell compartment predict durable, treatment-free DORIS remission after CD19 CAR-T?

Early cohorts describe naive-dominant B-cell return, but phenotype alone does not establish restored tolerance.

Evidence design
Proposed research design
Directness
Explicit inference
Inspect the evidence

Topic: CAR-T

Evidence that motivates the question

Boundary or counterevidence

Testable prediction

Participants with sustained remission will show durable depletion of pretreatment autoreactive clonotypes and lower functional autoreactivity after B-cell return.

Would weaken or falsify it

Durable remission is unrelated to B-cell repertoire or autoreactivity, or relapse occurs despite a persistently naive-dominant repertoire.

Proposed study

Embed blinded longitudinal BCR sequencing, antigen-reactivity assays, tissue-aware sampling where feasible, and standardized DORIS assessments in prospective CAR-T cohorts.

Safety boundary

This proposes measurement within regulated trials; it does not propose adding an intervention or changing patient treatment.

Site hypothesisVerified 2026-07-10

Is there an EBV-high mechanistic subgroup of SLE in which EBV-positive antigen-presenting B cells precede and predict flares?

The 2025 study supplies a candidate cell state, while existing epidemiology cannot establish temporal order.

Evidence design
Proposed research design
Directness
Explicit inference
Inspect the evidence

Topic: EBV

Evidence that motivates the question

Boundary or counterevidence

Testable prediction

A validated EBV-positive B-cell signature will rise before clinical activity in a reproducible subgroup and add predictive information beyond medication and baseline activity.

Would weaken or falsify it

The signature does not replicate, follows rather than precedes activity, or is explained by immunosuppression and generalized immune activation.

Proposed study

Prospective multi-center flare cohort with repeated EBV-cell-state, viral-load, medication, serologic, and disease-activity measurements using a preregistered analysis.

Safety boundary

Observational study only; the hypothesis does not support antiviral or vaccine treatment recommendations.

Site hypothesisVerified 2026-07-10

Does a reproducible microbiome state modify lupus activity or treatment response after accounting for medication, diet, geography, and baseline disease?

Human association, strain-specific immune responses, and mouse transfer results make a modifier effect plausible, but confounding and reverse causation remain strong alternatives.

Evidence design
Proposed research design
Directness
Explicit inference
Inspect the evidence

Topic: Microbiome

Evidence that motivates the question

Boundary or counterevidence

Testable prediction

A prespecified functional microbial signature will precede disease change and improve prediction in independent cohorts after major confounders are measured.

Would weaken or falsify it

The signature fails external replication or disappears after adjustment for medication, diet, geography, and disease activity.

Proposed study

Prospective multi-region cohort using shotgun metagenomics, metabolomics, standardized medication/diet capture, repeated activity outcomes, and external validation before any intervention trial.

Safety boundary

No FMT or live-organism regimen is proposed until observational signals replicate and a regulated safety protocol exists.

Site hypothesisVerified 2026-07-10

Does elimination of EBV-positive B-cell clones mediate any portion of clinical response after CD19 CAR-T?

CD19 CAR-T and EBV-positive B-cell biology create a plausible connection that neither source directly tests.

Evidence design
Proposed research design
Directness
Explicit inference
Inspect the evidence

Topic: Cross-cutting

Evidence that motivates the question

Boundary or counterevidence

Testable prediction

Reduction in validated EBV-positive autoreactive B-cell measures will statistically mediate later clinical response beyond total B-cell depletion.

Would weaken or falsify it

Clinical response is unrelated to EBV-positive clone burden or occurs without measurable change in that compartment.

Proposed study

Preplanned mechanistic substudy in existing CAR-T trials with baseline and serial EBV-positive B-cell assays; no additional therapeutic arm.

Safety boundary

This is a biomarker study, not a rationale to seek CAR-T or add antiviral treatment.

Site hypothesisVerified 2026-07-10

Do longitudinal blood and organ-specific interferon states predict domain-level disease change and response better than a one-time binary blood signature?

Interferon-inducible expression is common but heterogeneous, and phase 3 subgroup results do not validate baseline high-versus-low status as a sufficient treatment-selection rule.

Evidence design
Proposed research design
Directness
Explicit inference
Inspect the evidence

Topic: Type I interferon

Evidence that motivates the question

Boundary or counterevidence

Testable prediction

A preregistered continuous, longitudinal model combining blood interferon activity with organ-proximal measures will improve out-of-sample prediction of domain-specific activity and treatment response over baseline binary signature status and clinical covariates alone.

Would weaken or falsify it

The multi-compartment longitudinal model fails external validation, adds no clinically meaningful discrimination or calibration, or mainly captures infection, medication exposure, and concurrent rather than future activity.

Proposed study

Embed repeated standardized blood assays and optional clinically obtained tissue profiling in prospective multi-center SLE cohorts and interferon-pathway trials, with organ-specific outcomes, medication and infection capture, locked thresholds, and external validation.

Safety boundary

Biomarker measurement only; results must not direct treatment until prospectively validated and clinically reviewed.

Site hypothesisVerified 2026-07-10

Do BAFF dynamics and the composition of reconstituting B-cell compartments jointly predict relapse after distinct B-cell-directed therapies?

BAFF measurements are heterogeneous, BAFF inhibition is clinically active, and ligand blockade, antibody depletion, and CD19 CAR-T perturb the B-cell system at different depths.

Evidence design
Proposed research design
Directness
Explicit inference
Inspect the evidence

Topic: B cell / BAFF

Evidence that motivates the question

Boundary or counterevidence

Testable prediction

Within each prespecified treatment class, a model combining serial free BAFF, receptor occupancy, B-cell subset abundance, repertoire autoreactivity, and tissue-aware measures will predict standardized relapse outcomes better than total peripheral B-cell count alone.

Would weaken or falsify it

The combined measures do not replicate across cohorts, fail to precede relapse, or lose predictive value after adjustment for disease severity, background therapy, infection, and depletion depth.

Proposed study

A harmonized observational substudy across existing BAFF-inhibitor, anti-CD20, and regulated CAR-T cohorts using identical sampling windows and outcome definitions; analyze each treatment class separately before any cross-class comparison.

Safety boundary

Observational measurements only; the hypothesis does not support combining, switching, timing, or dosing B-cell-directed treatments.

Submission standard

A hypothesis is not finished until it names its failure conditions.

The lab is designed to resist attractive stories that cannot be falsified.

  1. Define the populationSpecify phenotype, organ involvement, treatment context, and major exclusions.
  2. Lock the variablesName exposure, comparator, outcome, time horizon, and expected direction.
  3. Plan for alternativesList confounding, reverse causation, selection, measurement, and model risks.
  4. State the stopping ruleSay what evidence would downgrade or retire the hypothesis.