Competing Risks Censoring Correction for Immunogenicity -- Anti-Drug Antibodies as Interval-Censored Competing Risk
Fixing a hidden flaw in drug safety testing: fast-failing proteins mask their immune risks until it's too late.
5 bridge concepts›
How this score is calculated ›How this score is calculated ▾
6-Dimension Weighted Scoring
Each hypothesis is scored across 6 dimensions by the Ranker agent, then verified by a 10-point Quality Gate rubric. A +0.5 bonus applies for hypotheses crossing 2+ disciplinary boundaries.
Is the connection unexplored in existing literature?
How concrete and detailed is the proposed mechanism?
How far apart are the connected disciplines?
Can this be verified with existing methods and data?
If true, how much would this change our understanding?
Are claims supported by retrievable published evidence?
Composite = weighted average of all 6 dimensions. Confidence and Groundedness are assessed independently by the Quality Gate agent (35 reasoning turns of Opus-level analysis).
RQuality Gate Rubric
0/10 PASS
| Criterion | Result |
|---|---|
| ABC Structure | true |
| Test Protocol | true |
| Counter-Evidence | true |
| Novelty | true |
| Precision | true |
| Groundedness Adequate | true |
| Mechanism | true |
| Confidence | true |
| Falsifiable | true |
| Claim Verification | true |
Claim Verification
Empirical Evidence
How EES is calculated ›How EES is calculated ▾
The Empirical Evidence Score measures independent real-world signals that converge with a hypothesis — not cited by the pipeline, but discovered through separate search.
Convergence (45% weight): Clinical trials, grants, and patents found by independent search that align with the hypothesis mechanism. Strong = direct mechanism match.
Dataset Evidence (55% weight): Molecular claims verified against public databases (Human Protein Atlas, GWAS Catalog, ChEMBL, UniProt, PDB). Confirmed = data matches the claim.
When pharmaceutical companies engineer therapeutic proteins — like the antibody drugs used to treat cancer or autoimmune diseases — one of their biggest fears is that the patient's immune system will recognize the drug as foreign and mount an attack against it. This reaction, called immunogenicity, can render a drug ineffective or even dangerous. To test for it, researchers measure whether patients develop 'anti-drug antibodies' (ADA) that neutralize the treatment. Meanwhile, a completely separate mathematical field called survival analysis has spent over 200 years developing tools to figure out *when* and *why* things fail — originally for actuarial tables, now used everywhere from clinical trials to engineering. This hypothesis connects those two worlds to expose a sneaky statistical trap. Some protein drugs fail quickly for entirely different reasons — they get chewed up by enzymes, or they clump together and get cleared from the body before the immune system even has a chance to notice them. The immune system needs at least 5-7 days to mount its first response. So if a drug disappears in 2 days, it never triggers a detectable immune reaction — not because it's safe, but because it didn't stick around long enough to be caught. This creates a systematic illusion: proteins that fail fast *look* less immunogenic than they really are. The hypothesis proposes using a specialized statistical model — designed to handle exactly this kind of overlapping, competing failure scenario — to estimate what the *true* immune risk would be, if the drug actually stayed in the body long enough. Why does this matter? Because drug engineers are constantly working to make protein drugs last longer in the body — adding chemical modifications or fusing them to other proteins to extend their lifespan. If a drug had hidden immune risk that was masked by its short lifespan, making it last longer could suddenly unleash that immune reaction in patients. The hypothesis calls this 'ADA unmasking' — and it's a potential clinical disaster hiding in plain sight.
This is an AI-generated summary. Read the full mechanism below for technical detail.
Why This Matters
If confirmed, this framework could fundamentally change how biopharmaceutical companies and regulators assess the safety of protein-based drugs before and during clinical trials. Drug designers could use the model to flag proteins with high 'latent immunogenicity' early in development, before investing in expensive half-life extension strategies that might backfire in patients. Regulatory agencies like the FDA and EMA could require corrected immunogenicity estimates that account for competing failure modes, raising the bar for what counts as a safe drug. This is especially urgent as AI-designed proteins — an exploding field capable of generating millions of novel candidates — enter clinical pipelines, since those proteins may have unpredictable immune profiles that current testing would systematically underestimate. The hypothesis is testable with existing clinical datasets by comparing ADA rates before and after half-life extension for the same protein sequences, making it worth pursuing now.
Mechanism
Among the five competing failure modes, immunogenicity has a unique temporal structure with a minimum biological latency (5-7 days primary, 2-3 days secondary ADA response). ADA is detectable only at discrete sampling times, making it interval-censored. Crucially, proteins that are rapidly cleared by proteolysis or aggregation NEVER REACH the immunogenicity window -- they fail before the immune system can respond. This creates informative censoring: rapid non-immune clearance systematically biases observed ADA rates downward.
An interval-censored competing risks model (Sun 2006) jointly models all failure modes while correctly handling the interval censoring of ADA onset, estimating the "latent immunogenicity" -- the ADA rate that WOULD be observed if the protein survived long enough. This is critical for design decisions: a protein with high latent immunogenicity but low observed ADA (because it fails fast from other causes) will become a clinical problem if designers successfully extend its half-life.
Supporting Evidence
Key strength: Identifies systematic bias in immunogenicity assessment with regulatory implications. Predictions: Compare ADA rates for the same protein sequence with and without half-life extension (PEGylation or Fc fusion). The competing risks model predicts ADA unmasking at a rate predicted by the latent immunogenicity estimate.. Groundedness: 7/10. Claims verified: 5, failed: 0.. Application pathway: diagnostic (Biopharmaceutical immunogenicity / regulatory science)
How to Test
Compare ADA rates for the same protein sequence with and without half-life extension (PEGylation or Fc fusion). The competing risks model predicts ADA unmasking at a rate predicted by the latent immunogenicity estimate.
Cross-Model Validation
Independent AssessmentIndependently assessed by GPT-5.4 Pro and Gemini 3.1 Pro for triangulation. Assessed independently by two external models for triangulation.
Other hypotheses in this cluster
The Dominant Competing Risk Theorem -- Optimizing One Failure Mode Provably Accelerates Another
PASSFix one way a protein drug breaks, and you mathematically guarantee another weakness gets worse.
Competing-Risk Cumulative Incidence Functions as a Unified Protein Therapeutic Lifetime Predictor
PASSA survival statistics framework borrowed from actuaries could predict exactly how—and when—engineered protein drugs will break down in the body.
Nelson-Aalen Cumulative Hazard Decomposition Reveals Hidden Failure Modes in Accelerated Stability Studies
CONDITIONALSplitting protein drug degradation into its hidden failure modes could make shelf-life predictions far more accurate.
Related hypotheses
Gaussian Mixture Model Analysis of Cryo-EM OMV Populations Distinguishes Biogenesis Pathways in P. aeruginosa
PASSAI-powered microscopy could reveal how bacteria decide what to pack into their tiny 'mail packages'.
Abiotic vs Enzymatic PLOOH Regioselectivity as Chemical Fossil of Antioxidant Evolution
PASSThe chaotic chemistry of ancient iron reactions may have driven evolution of the precise enzymes that now control cell death.
Pyocyanin-GPX4-Ferroptosis Bidirectional Axis
PASSBacteria may hack their own iron supply by triggering a specific type of cell death in human lung cells.
Can you test this?
This hypothesis needs real scientists to validate or invalidate it. Both outcomes advance science.