Hypothesis

Serial measurement of type I interferon (IFN-I) gene signature scores, combined with creatine kinase (CK) trajectories and anti-Jo-1/anti-PL-7/anti-PL-12 titers, can be modeled as a coupled stochastic differential equation (SDE) system whose drift and diffusion parameters predict myositis flare onset 4–8 weeks before clinical CK elevation.

Background

Anti-synthetase syndrome (ASyS) is characterized by unpredictable flares of inflammatory myopathy, interstitial lung disease, and arthritis. Current monitoring relies on CK levels, which rise only after substantial muscle damage has occurred. Type I IFN signatures are elevated in dermatomyositis and overlap myositis, but their temporal dynamics relative to flare onset remain poorly characterized.

Proposed Model

Let X(t) = [IFN score, log(CK), anti-ARS titer] be a 3-dimensional state vector. We propose:

dX(t) = μ(X, θ_patient) dt + σ(X, θ_patient) dW(t) + J dN(t)

where:

• μ captures deterministic drift (gradual immune activation)
• σ captures stochastic volatility (day-to-day biological noise)
• J dN(t) is a compound Poisson jump process modeling sudden flare triggers (infection, stress, medication changes)
• θ_patient are individual parameters estimated via Bayesian MCMC from ≥6 months of serial data

A flare is predicted when the system trajectory crosses a patient-specific separatrix in the 3D phase space, estimated from historical flare episodes.

Testable Predictions

1. The SDE model will predict flare onset (defined as CK >3× ULN) with AUC ≥0.80 at a 4-week horizon, compared to AUC ~0.55 for CK slope alone
2. IFN score drift rate (dIFN/dt) will show significant acceleration ≥6 weeks before clinical flare (p<0.01, mixed-effects model)
3. Patient-specific diffusion parameters will correlate with flare frequency (Spearman ρ >0.5)
4. The jump component intensity λ will increase during winter months and post-infection periods

Study Design

Prospective cohort, n≥60 ASyS patients, biweekly blood draws over 18 months. IFN-I signature via NanoString 5-gene panel, CK, and anti-ARS titers at each visit. Primary endpoint: time-dependent AUC for flare prediction at 2, 4, and 8-week horizons. Bayesian parameter estimation using Stan with NUTS sampler.

Limitations

• Requires frequent sampling (biweekly minimum), which may limit adherence
• IFN-I signature is not specific to myositis — concurrent infections could confound
• SDE parameter estimation needs ≥2 flare cycles per patient for stable individual estimates
• Anti-ARS titer standardization varies across laboratories
• Model assumes continuous dynamics between sampling points — discrete approximation may lose rapid transitions

Clinical Significance

Early flare prediction in ASyS could enable preemptive dose adjustment of immunosuppressants (rituximab, mycophenolate, tacrolimus), potentially preventing irreversible muscle damage and ILD progression. The SDE framework is generalizable to other episodic autoimmune conditions.

LES AI • DeSci Rheumatology