STRC Research

STRC mRNA-LNP PKPD Multi-Dose Schedule

7 min read

STRC mRNA-LNP PK/PD Multi-Dose Schedule

Three-compartment PK/PD integration across dose-interval × per-dose × modification × LNP-targeting. Extends the single-dose STRC mRNA Therapy Hypothesis into a repeat-dosing schedule model. Two independent findings: (1) per-OHC pharmacology is solved at modest dose — m1ψ mRNA at 200 mol/OHC intracellular Q6W keeps transfected-OHC STRC trough ≥ 2.17× (1,800 mol/OHC/yr intracellular = 90,000 mol/OHC/yr extracellular at 2% endosomal escape); (2) LNP delivery is the real ceiling — cochlea-mean STRC fold is bounded by eff × 3 + (1−eff) ≈ 1.4× even at 20% OHC targeting, so no PK/PD regimen can push the cochlea-wide average ≥ 2×. The hypothesis decouples into a pharmacology sub-problem (solved) and a delivery sub-problem (open; requires LNP tropism > 50% to clear cochlea-mean threshold).

Method

Compartments: RBM24 mRNA m → RBM24 protein p → STRC fold f.

dm/dt = pulse(t; D, T) − k_m · m
dp/dt = k_tr · m − k_p · p
df/dt = k_f · [ 1 + (3−1) · Hill(p) − f ]
Hill(p) = p² / (200² + p²)

Parameters inherited from single-dose calibration in rbm24_mrna_dose_response_results.json:

ParameterValueSource
K_Hill (K_M for RBM24→STRC boost)200 (rel. units)dose-response fit
Hill n2dose-response fit
max_boost3.0×dose-response fit
threshold_fold2.0×pharmacology go/no-go
mRNA t½ (unmod, OHC)4 hmrna_stability_cochlear_results.json
mRNA t½ (m1ψ, OHC)12 hidem
RBM24 protein t½2 didem
STRC protein t½14 didem (dominant eigenvalue)
LNP endosomal escape2%literature

Pulse: intracellular dose D (mol/OHC) injected as Δm = D at t = 0, T, 2T, .... Extracellular LNP burden = D / 0.02.

Population layer: transfection efficiency eff_ohc. Cochlea-mean fold = eff × f_tx + (1 − eff) × 1. Three scenarios:

Targetingeff_ohcSource
Untargeted0.8%96/12000 OHC (standard LNP, literature)
Cochlear-tropic5%hypothetical next-gen LNP
OHC-targeted20%hypothetical ligand-conjugated LNP

Sweep: 6 intervals × 5 doses × 2 modifications × 3 targetings = 180 regimens. 365-day simulation, steady-state trough/peak measured in last 30% of trajectory. LSODA with breakpoints at dose times. Deterministic. Runtime < 2 min on one CPU.

Results — per-OHC PK/PD (targeting-independent)

Transfected-OHC steady-state fold at minimum dose (D=200 mol/OHC) vs interval:

IntervalUnmod t½=4h troughm1ψ t½=12h troughm1ψ therapeutic?
Q2W (14 d)3.00×3.00×yes, saturated
Q3W (21 d)3.00×3.00×yes, saturated
Q4W (28 d)2.93×2.99×yes
Q6W (42 d)1.92×2.17×m1ψ yes, unmod no @ 200
Q8W (56 d)1.43×1.55×no
Q12W (84 d)1.10×1.13×no

Interval ceiling ≈ 6 weeks. Beyond Q6W, STRC protein half-life (14 d = 0.33 × 42 d, 0.25 × 56 d) cannot bridge the gap between RBM24 pulses, trough collapses below 2×.

Minimum annual burden per-OHC (transfected)

ModificationMin intervalMin dose/pulseDoses/yrIntra/yrExtra/yr (÷0.02)
m1ψ (best)Q6W200 mol91,800 mol90,000 mol
unmodQ4W200 mol132,600 mol130,000 mol
unmodQ6W500 mol94,500 mol225,000 mol

m1ψ vs unmod gap grows at long intervals. At Q2W both saturate; at Q6W m1ψ is 2.5× more dose-efficient than unmod (200 vs 500 mol minimum). m1ψ gets more translation per mRNA molecule by extending intracellular dwell — directly relevant at clinically meaningful dose intervals.

Results — LNP delivery is the ceiling

Cochlea-mean fold is bounded above by eff × max_boost + (1 − eff) × baseline = eff × 3 + (1 − eff):

LNP targetingeff_ohcTheoretical ceiling (cochlea-mean)Model-observed max
Untargeted0.8%1.016×1.016×
Cochlear-tropic5%1.10×1.10×
OHC-targeted20%1.40×1.40×

At 20% OHC targeting, cochlea-mean fold cannot exceed 1.4× no matter the dose. The remaining 80% of OHCs that never receive LNP dilute the response below any cochlea-wide threshold ≥ 2×. To breach 2× cochlea-mean: eff × 3 + (1−eff) ≥ 2 → eff ≥ 50%. No published LNP achieves this cochlear tropism.

Interpretation

  1. The mRNA therapy hypothesis has two bottlenecks, and the pharmacology one is effectively solved. The three-compartment model shows that once a transfected OHC receives ≥ 200 mol/dose Q6W of m1ψ-modified RBM24 mRNA, STRC boost stays saturated near 2.2–2.8×. STRC-protein t½ (14 d) does most of the work smoothing the pulse pharmacokinetics — the cochlea is a transcript integrator.

  2. Delivery is the only thing standing between “interesting ODE” and “therapy.” Moving from 20% to 50% OHC targeting is a pure LNP-chemistry problem: cochlear-tropic lipids (SORT-LNP, GalNAc-equivalent cochlear ligands), round-window delivery geometry, ionic envelope of endolymph. No amount of dose optimization fixes it.

  3. Per-region rescue, not cochlea-mean, is the clinical endpoint. STRC is a structural protein acting locally at each hair bundle. Cochlea-mean is the wrong metric for a structural protein: 20% OHC rescue could restore ~20% of audiogram coverage (one tonotopic band at a time), even though cochlea-mean fold is only 1.4×. The “need ≥ 50% coverage” framing applies only if uniform rescue is the goal; if targeted-frequency rescue is acceptable, 20% is already therapeutic locally. An audiogram-level endpoint model needs to be built to make this rigorous.

  4. For Misha (compound het: 98 kb paternal Δ + c.4976A>C E1659A maternal): the RBM24 strategy works only on the maternal (hypomorphic) allele — the paternal allele has no pre-mRNA to splice-correct. So mRNA-RBM24 delivery at best doubles expression from one allele, giving ~1× physiological STRC (from 0.5× hemizygous baseline). This is likely not enough on its own; pairs well with STRC Pharmacochaperone Virtual Screen E1659A (quantity × affinity additive) or with Strategy B (full-length STRC mRNA, not RBM24).

Limitations

  • Single compartment per OHC. No model of mRNA-LNP trafficking (endosomal escape kinetics), only a lumped 2% multiplier. A proper model would integrate vesicle-escape rate into k_m explicitly.
  • No extracellular perilymph compartment. LNP clearance from perilymph is modeled as instantaneous uptake by the 96/12000 = 0.8% fraction of OHCs that get any dose. A distribution model (e.g., Stokes-Einstein diffusion through round window + basilar membrane barrier) would shift transfection dynamics.
  • Linear cochlea-mean is not the clinical endpoint. A tonotopic audiogram model should replace the mean with per-frequency-band rescue summing.
  • No immune/anti-PEG response; LNP readministration may trigger anti-PEG antibodies that shift PK after 2-3 doses in vivo. Relevant for Q2W/Q3W regimens specifically.
  • Hill parameters (K=200, n=2, max_boost=3) from fit to single-dose ODE; independent biochemical validation would improve confidence.
  • Does not model Strategy B (full-length STRC mRNA). That strategy bypasses the RBM24 splicing layer entirely — different ODE, same delivery bottleneck.

Next steps

  1. Audiogram-rescue model: given eff_ohc = 20% and per-OHC trough = 2.17×, what tonotopic frequency bands recover by how many dB? Couples this PK/PD model to STRC Stereocilia Bundle Mechanics Model and Misha’s audiogram.
  2. LNP tropism literature scan: map published cochlear LNP platforms and their eff_ohc; identify a 50% target candidate.
  3. Strategy B PK/PD: repeat this analysis with full-length STRC mRNA (Strategy B, bypasses RBM24) — simpler model (mRNA → STRC protein directly), same delivery ceiling.
  4. Anti-PEG model: add immunogenicity clearance to k_m at doses 3+ to bound realistic chronic-dosing horizons.
  5. Stochastic transfection: each OHC’s LNP uptake is a Bernoulli event. A Gillespie simulation could answer how variable the per-OHC fold-trough is across a cochlear population.

Replication

cd ~/STRC/models
/opt/miniconda3/bin/python3 mrna_lnp_pkpd_integration.py
# outputs: mrna_lnp_pkpd_integration.json

Files / Models

  • ~/STRC/models/mrna_lnp_pkpd_integration.py — 3-compartment ODE + dose-pulse sweep
  • ~/STRC/models/mrna_lnp_pkpd_integration.json — all 180 regimens, minimum-therapeutic table, coverage ceiling, exemplar traces
  • ~/STRC/models/mrna_stability_cochlear_results.json — half-life priors (input)
  • ~/STRC/models/rbm24_mrna_dose_response_results.json — Hill fit priors (input)

Connections

Graph View

Backlinks