Skip to contents

Translational TfR-mediated brain delivery of antibodies (Muliaditan 2025)

Source: vignettes/articles/Muliaditan_2025_mab_mpbpk.Rmd
Muliaditan_2025_mab_mpbpk.Rmd

Model and source

  • Citation: Muliaditan M, van Steeg TJ, Avery LB, Sun W, Hammond TR, Hijdra D, Choi SL, Pillai N, Leksa NC, Mavroudis PD. Translational minimal physiologically based pharmacokinetic model for transferrin receptor-mediated brain delivery of antibodies. mAbs. 2025;17(1):2515414.
  • DOI: https://doi.org/10.1080/19420862.2025.2515414
  • PMID: 40568753

The paper develops a minimal physiologically based pharmacokinetic (mPBPK) model for monoclonal antibody (mAb) delivery to the brain via transferrin-receptor (TfR) mediated transcytosis. The mPBPK framework combines (a) the Bloomingdale 2017 mAb mPBPK structure (plasma, lumped tissue vascular / endosomal / interstitial spaces, brain vascular, blood-brain barrier (BBB) endosome, brain ISF, blood-CSF barrier (BCSFB) endosome, CSF, lymph, with FcRn-mediated recycling of intact antibody from each endosomal compartment) and (b) Chang 2022’s TfR binding sites at the plasma, luminal BBB, luminal BCSFB, abluminal BBB (brain ISF), abluminal BCSFB (CSF), and neuronal compartments. The bsAb-TfR complex is transcytosed across the brain barriers at a fixed rate ktrans = 6 h^-1 and is empirically degraded at rate kint (modelled as a binary fast / slow mixture).

The implementation in this package ships as two paired model files:

  • Muliaditan_2025_mab_mpbpk_nhp – the primary fit, parameterised for the cynomolgus monkey (6.2 kg reference body weight, Bloomingdale 2017 NHP physiology in Supp Table S1). 26 ODE compartments.
  • Muliaditan_2025_mab_mpbpk_human – the paper’s translation of the NHP fit to a 70-kg human, with kint allometrically scaled by (70/6.2)^(-0.25) = 0.546 and uTFR0_BCSFB recalibrated 3-fold higher than the NHP estimate.

Both files share the same ODE structure, the same residual error structure, and the same per-compound TfR binding parameter conventions (kon_T, koff_T, mw_kda set by the user from the Muliaditan 2025 Supplementary Table S2 for each simulated antibody). The vignette covers both files in one narrative.

Population

The NHP training dataset (paper Table 1) contains 395 plasma, 81 CSF, and 102 brain mean concentrations digitised from eight literature preclinical studies covering 7 non-TfR mAbs and 10 anti-TfR bispecific antibodies (bsAbs) with TfR binding affinities ranging from KD,TfR = 36 to 1900 nM. The cohort included compounds from Atwal 2011 (control IgG), Edavettal 2022 (TfR-J wild type and mutant, TfR-K mutant), Fjord-Larsen 2021 (Lu-AF82422), Grimm 2023 (gantenerumab and trontinemab), Kanodia 2016 / Chang 2022 (anti-TfR/BACE1 at three affinities), Kariolis 2020 (ATV35.21.16:BACE1 and a control IgG), Yadav 2017 (anti-BACE1), and Yu 2014 (anti-TfR1-BACE1 and anti-TfR2-BACE1, plus a control IgG). The animals received single IV doses of 1, 2, 3, 10, 30, or 50 mg/kg, except for Yadav 2017 (four weekly infusions) and Edavettal 2022 RD arm (repeated dosing every three weeks). Inter-individual variability was not evaluated because the dataset was largely mean profiles digitised from published figures.

For the clinical validation (paper Figure 5) the paper used CSF concentrations and plasma exposure metrics following single ascending doses of trontinemab (0.1, 0.4, 1.2, 3.6, 7.2 mg/kg IV) in healthy adult subjects derived from the Phase 1 trial NCT04023994 (Grimm 2023).

The population details are available programmatically:

mod_nhp_meta <- rxode2::rxode(nlmixr2lib::readModelDb("Muliaditan_2025_mab_mpbpk_nhp"))
str(mod_nhp_meta$population)
#> List of 11
#>  $ species      : chr "cynomolgus monkey"
#>  $ n_subjects   : int NA
#>  $ n_studies    : int 8
#>  $ n_compounds  : int 17
#>  $ n_plasma_obs : int 395
#>  $ n_csf_obs    : int 81
#>  $ n_brain_obs  : int 102
#>  $ weight_range : chr "Bloomingdale 2017 6.2-kg cynomolgus reference subject for the physiology"
#>  $ disease_state: chr "healthy non-human primate (cynomolgus monkey)"
#>  $ dose_range   : chr "1-100 mg/kg IV bolus or IV infusion (training data: 2, 10, 30, 50 mg/kg single IV; Edavettal 2022 also includes"| __truncated__
#>  $ notes        : chr "Training dataset (paper Table 1) covers 7 non-TfR mAbs (Control IgG, anti-Tau-IgG, Lu-AF82422, gantenerumab, an"| __truncated__

Source trace

The per-parameter origin is recorded as an in-file comment next to every ini() entry in the model files. The table below collects the parameter values for review.

Parameter Value (NHP) Source
TfRpt total plasma TfR 1672 nM (RSE 38%) Table 2
uTFR0_BBB baseline luminal BBB TfR 175 nM (RSE 34%) Table 2
uTFR0_BCSFB baseline luminal BCSFB TfR 0.256 nM (RSE 47%) Table 2
TfRtotn neuronal TfR 559 nM (FIXED, Chang 2022) Table 2
ktrans BBB / BCSFB transcytosis 6 h^-1 (FIXED, Chang 2022) Table 2
kdeg_uTfRBBB luminal BBB TfR degradation 20 h^-1 (FIXED) Table 2
kdeg_uTfRBCSFB luminal BCSFB TfR degradation 20 h^-1 (FIXED) Table 2
Fraction POP1 0.437 (RSE 30%) Table 2; mapped to MIX_FAST_ELIM canonical
kint POP1 fast complex internalization 0.0329 h^-1 (RSE 14%) Table 2
kint POP2 slow complex internalization 0.0125 h^-1 (RSE 8.4%) Table 2
krec_uTfR = kint (FIXED) Table 2
FAC_BPRED brain prediction correction 0.05 (FIXED) Table 2
FACQ_BECF ISF -> CSF correction 0.00814 (RSE 29%) Table 2
sigma^2 plasma (additive on log scale) 0.256 (RSE 26%) Table 2
sigma^2 CSF 0.893 (RSE 24%) Table 2
sigma^2 brain 0.426 (RSE 38%) Table 2
Vp, VTv, VTe, VTi, VBv, V_BE_BBB, V_BE_BCSFB, VBi, VCSF, VL NHP / human physiology Supp Table S1
QT, QB, LT, LB, Q_BECF, Q_BCSF NHP / human physiology Supp Table S1
kon_FcRn, koff_FcRn, Kdeg (endosomal), FR, FR_B, FcRn baseline NHP / human physiology Supp Table S1
sigma_BBB, sigma_BCSFB, sigma_Tv, sigma_TL, sigma_ISF, sigma_CSF NHP / human physiology Supp Table S1
CLUPT, CLUPBBB, CLUPBCSFB, k_CLUPT, k_CLUPB NHP / human physiology Supp Table S1
Human kint (POP1) 0.0179 h^-1 (= 0.0329 * 0.546) Methods (allometric exponent -0.25)
Human kint (POP2) 0.00683 h^-1 (= 0.0125 * 0.546) Methods (same allometric scaling)
Human uTFR0_BCSFB 0.768 nM (= 0.256 * 3) Results (3-fold recalibration)
Per-compound kon_T, koff_T, MW per antibody Supp Table S2
ODE system 26 equations Supplementary “Model equations” section

Helper: dose conversion mg -> nmol

The model uses concentrations in nM and amounts in nmol throughout. A clinical dose entered in mg is converted to nmol inside the model via f(central) = 1000 / mw_kda, so the user supplies amt in mg of antibody and mw_kda in kDa.

# 10 mg/kg in a 6.2 kg NHP
nhp_wt_kg   <- 6.2
nhp_dose_mg <- 10 * nhp_wt_kg   # = 62 mg trontinemab

# Trontinemab in human: paper used 0.1, 0.4, 1.2, 3.6, 7.2 mg/kg single doses
human_wt_kg     <- 70
trontinemab_mw  <- 194   # kDa; Muliaditan 2025 Table S2

# Trontinemab TfR binding parameters (paper Table S2)
tron_nhp_kon   <- 0.21168     # nM^-1 h^-1 (NHP)
tron_nhp_koff  <- 52.56       # h^-1      (NHP)
tron_hum_kon   <- 1.0548      # nM^-1 h^-1 (human)
tron_hum_koff  <- 138.24      # h^-1      (human)

Simulation 1 – trontinemab in cynomolgus monkey (replicates Figure 3 trontinemab panel)

A single 10 mg/kg IV bolus is simulated for 28 days (672 h). The slow kint subpopulation (MIX_FAST_ELIM = 0) is used as the typical-value prediction; the figure overlays the fast and slow subpopulations to show the kint sensitivity that drives the mixture-model fit.

mod_nhp <- nlmixr2lib::readModelDb("Muliaditan_2025_mab_mpbpk_nhp")
mod_nhp_typ <- rxode2::zeroRe(mod_nhp)
#> Warning: No omega parameters in the model

times <- seq(0.25, 672, length.out = 240)

ev_nhp <- rxode2::et(time = 0, amt = nhp_dose_mg, cmt = "central") |>
  rxode2::et(times, cmt = "Cc")

sim_nhp_pop1 <- rxode2::rxSolve(
  mod_nhp_typ,
  events = ev_nhp,
  params = c(MIX_FAST_ELIM = 1,
             lkon_t  = log(tron_nhp_kon),
             lkoff_t = log(tron_nhp_koff),
             mw_kda  = trontinemab_mw)
) |> as.data.frame()

sim_nhp_pop2 <- rxode2::rxSolve(
  mod_nhp_typ,
  events = ev_nhp,
  params = c(MIX_FAST_ELIM = 0,
             lkon_t  = log(tron_nhp_kon),
             lkoff_t = log(tron_nhp_koff),
             mw_kda  = trontinemab_mw)
) |> as.data.frame()

sim_nhp <- bind_rows(
  sim_nhp_pop1 |> mutate(subpop = "POP1 (fast kint)"),
  sim_nhp_pop2 |> mutate(subpop = "POP2 (slow kint)")
) |>
  filter(time > 0) |>
  pivot_longer(c(Cc, Ccsf, Cbrain),
               names_to = "Matrix", values_to = "Conc_nM")
ggplot(sim_nhp, aes(time, pmax(Conc_nM, 1e-6), colour = subpop, linetype = subpop)) +
  geom_line() +
  facet_wrap(~ factor(Matrix, levels = c("Cc", "Ccsf", "Cbrain"),
                       labels = c("Plasma (total)", "CSF (unbound)", "Brain homogenate")),
             scales = "free_y") +
  scale_y_log10() +
  scale_x_continuous(breaks = seq(0, 672, 168)) +
  labs(x = "Time (h)", y = "Concentration (nM, log)",
       colour = NULL, linetype = NULL,
       title = "Trontinemab 10 mg/kg single IV in cynomolgus monkey")
Trontinemab 10 mg/kg in NHP: simulated plasma total drug (Cc), CSF unbound (Ccsf), and whole-brain homogenate (Cbrain) concentrations vs. time. POP1/POP2 are the two kint subpopulations from the mixture model. Replicates Figure 3 trontinemab panel of Muliaditan 2025.

Trontinemab 10 mg/kg in NHP: simulated plasma total drug (Cc), CSF unbound (Ccsf), and whole-brain homogenate (Cbrain) concentrations vs. time. POP1/POP2 are the two kint subpopulations from the mixture model. Replicates Figure 3 trontinemab panel of Muliaditan 2025.

Simulation 2 – non-TfR control mAb in NHP (sanity check)

Setting kon_T = 0 (encoded as a tiny placeholder 1e-12 in ini()) collapses all TfR-mediated dynamics. The simulated profiles should look like a generic linear-PK mAb in plasma with very low brain / CSF exposure (the FAC_BPRED = 0.05 brain-correction factor scales the brain prediction downward by 20-fold relative to the brain ISF concentration).

sim_control <- rxode2::rxSolve(
  mod_nhp_typ,
  events = ev_nhp,
  params = c(MIX_FAST_ELIM = 0,
             # Default ini() already encodes non-TfR control; show parameters
             # explicitly here for clarity.
             lkon_t  = log(1e-12),
             lkoff_t = log(1),
             mw_kda  = 150)   # generic IgG
) |> as.data.frame()

control_summary <- sim_control |>
  filter(time > 0) |>
  summarise(Cc_max = max(Cc), Ccsf_max = max(Ccsf), Cbrain_max = max(Cbrain))
knitr::kable(control_summary,
             caption = "Peak concentrations for a non-TfR control mAb at 10 mg/kg in NHP")
Peak concentrations for a non-TfR control mAb at 10 mg/kg in NHP
Cc_max Ccsf_max Cbrain_max
1195.233 1.924228 0.4767569

The brain and CSF peaks for the non-TfR control are much lower than the corresponding peaks for trontinemab (Simulation 1) – consistent with the paper’s observation that anti-TfR bsAbs achieve enhanced brain delivery relative to non-binders.

Simulation 3 – trontinemab in human, single ascending doses (replicates Figure 5) 

mod_hum <- nlmixr2lib::readModelDb("Muliaditan_2025_mab_mpbpk_human")
mod_hum_typ <- rxode2::zeroRe(mod_hum)
#> Warning: No omega parameters in the model

human_doses_mgkg <- c(0.1, 0.4, 1.2, 3.6, 7.2)
human_doses_mg   <- human_doses_mgkg * human_wt_kg

times_h <- seq(0.5, 168, length.out = 120)

build_human_events <- function(dose_mg, id_offset) {
  rxode2::et(time = 0, amt = dose_mg, cmt = "central") |>
    rxode2::et(times_h, cmt = "Cc") |>
    as.data.frame() |>
    mutate(id = id_offset + 1L,
           dose_mgkg = dose_mg / human_wt_kg)
}

events_human <- purrr::map2_dfr(
  human_doses_mg, seq_along(human_doses_mg) - 1L,
  build_human_events
)

sim_human <- rxode2::rxSolve(
  mod_hum_typ,
  events = events_human,
  params = c(MIX_FAST_ELIM = 0,
             lkon_t  = log(tron_hum_kon),
             lkoff_t = log(tron_hum_koff),
             mw_kda  = trontinemab_mw),
  keep = "dose_mgkg"
) |> as.data.frame()
#> Warning: multi-subject simulation without without 'omega'
sim_human_plot <- sim_human |>
  filter(time > 0) |>
  pivot_longer(c(Cc, Ccsf), names_to = "Matrix", values_to = "Conc_nM") |>
  mutate(dose_label = sprintf("%.1f mg/kg", dose_mgkg))

ggplot(sim_human_plot, aes(time, pmax(Conc_nM, 1e-6), colour = dose_label)) +
  geom_line() +
  facet_wrap(~ factor(Matrix, levels = c("Cc", "Ccsf"),
                       labels = c("Plasma (total)", "CSF (unbound)")),
             scales = "free_y") +
  scale_y_log10() +
  labs(x = "Time (h)", y = "Concentration (nM, log)",
       colour = "Dose",
       title = "Trontinemab single ascending dose in human")
Trontinemab single ascending IV doses (0.1-7.2 mg/kg) in healthy adult subjects: simulated plasma total drug Cc and CSF unbound Ccsf vs. time. Replicates Figure 5 of Muliaditan 2025.

Trontinemab single ascending IV doses (0.1-7.2 mg/kg) in healthy adult subjects: simulated plasma total drug Cc and CSF unbound Ccsf vs. time. Replicates Figure 5 of Muliaditan 2025.

Simulation 4 – predicted brain AUC vs TfR binding affinity (replicates Figure 4)

The paper’s Figure 4 sweeps KD,TfR from 0 to 3000 nM in NHP at single ascending doses 1-100 mg/kg and shows that dose-normalised brain AUC0-672h peaks in the 800-3000 nM range. This block reproduces the qualitative shape at a single 10 mg/kg dose.

kd_grid <- c(0, 30, 100, 300, 1000, 3000)   # nM
# Use the paper's average per-compound kon_T = 0.5193 nM^-1 h^-1, then
# derive koff_T from koff = KD * kon.
kon_T_typical <- 0.5193

sweep_brain_auc <- function(kd_nM) {
  if (kd_nM == 0) {
    kon  <- 1e-12
    koff <- 1
  } else {
    kon  <- kon_T_typical
    koff <- kd_nM * kon_T_typical
  }
  sim <- rxode2::rxSolve(
    mod_nhp_typ,
    events = rxode2::et(time = 0, amt = nhp_dose_mg, cmt = "central") |>
              rxode2::et(seq(2, 672, length.out = 120), cmt = "Cbrain"),
    params = c(MIX_FAST_ELIM = 0,
               lkon_t  = log(kon),
               lkoff_t = log(koff),
               mw_kda  = trontinemab_mw)
  ) |> as.data.frame()
  # Trapezoidal AUC over 0-672 h for brain homogenate concentration
  sim <- sim |> filter(!is.na(Cbrain)) |> arrange(time)
  auc <- sum(diff(sim$time) * (head(sim$Cbrain, -1) + tail(sim$Cbrain, -1)) / 2)
  data.frame(KD_nM = kd_nM, brain_AUC_nMh = auc)
}

kd_sweep <- purrr::map_dfr(kd_grid, sweep_brain_auc)
knitr::kable(kd_sweep, caption = "Brain AUC0-672h vs TfR KD (10 mg/kg NHP)")
Brain AUC0-672h vs TfR KD (10 mg/kg NHP)
KD_nM brain_AUC_nMh
0 149.3032
30 949.3885
100 2135.2927
300 3678.1496
1000 5024.8554
3000 4125.2488 
control_auc <- kd_sweep$brain_AUC_nMh[kd_sweep$KD_nM == 0]
ggplot(kd_sweep |> filter(KD_nM > 0),
       aes(KD_nM, brain_AUC_nMh)) +
  geom_line() +
  geom_point() +
  geom_hline(yintercept = control_auc, linetype = "dashed") +
  scale_x_log10() +
  labs(x = "KD,TfR (nM, log)",
       y = "Brain AUC0-672h (nM*h)",
       title = "Brain AUC vs TfR binding affinity in NHP")
Dose-normalised brain AUC0-672h vs TfR binding affinity in NHP at 10 mg/kg. Horizontal dashed line is the non-TfR control mAb. Replicates the qualitative finding of Figure 4 of Muliaditan 2025.

Dose-normalised brain AUC0-672h vs TfR binding affinity in NHP at 10 mg/kg. Horizontal dashed line is the non-TfR control mAb. Replicates the qualitative finding of Figure 4 of Muliaditan 2025.

PKNCA validation – human trontinemab plasma exposure

PKNCA is applied to plasma total drug Cc from Simulation 3 to compute Cmax, Tmax, AUC0-168h, AUC0-inf, and half-life per dose group, then compared against the published metrics in Muliaditan 2025 Figure 5c. The paper reports observed plasma AUC0-168h, AUC0-inf, and Cmax from the trontinemab Phase 1 trial (NCT04023994). Because the paper only shows these graphically (no per-dose tabulated table in the main text), the comparison here is qualitative.

sim_human_nca <- sim_human |>
  dplyr::filter(!is.na(Cc)) |>
  dplyr::select(id, time, Cc, dose_mgkg) |>
  dplyr::mutate(treatment = sprintf("%.1f mg/kg", dose_mgkg))

# Guarantee a time-zero row per (id, treatment); for IV bolus pre-dose Cc = 0.
sim_human_nca <- dplyr::bind_rows(
  sim_human_nca,
  sim_human_nca |> dplyr::distinct(id, treatment, dose_mgkg) |>
    dplyr::mutate(time = 0, Cc = 0)
) |>
  dplyr::distinct(id, treatment, time, .keep_all = TRUE) |>
  dplyr::arrange(id, treatment, time)

conc_obj <- PKNCA::PKNCAconc(sim_human_nca, Cc ~ time | treatment + id)

dose_df <- events_human |>
  dplyr::filter(evid == 1) |>
  dplyr::mutate(treatment = sprintf("%.1f mg/kg", dose_mgkg)) |>
  dplyr::select(id, time, amt, treatment)

dose_obj <- PKNCA::PKNCAdose(dose_df, amt ~ time | treatment + id)

intervals <- data.frame(
  start       = 0,
  end         = 168,
  cmax        = TRUE,
  tmax        = TRUE,
  auclast     = TRUE,
  aucinf.obs  = TRUE,
  half.life   = TRUE
)

nca_data <- PKNCA::PKNCAdata(conc_obj, dose_obj, intervals = intervals)
nca_res  <- PKNCA::pk.nca(nca_data)

nca_tbl <- as.data.frame(nca_res$result) |>
  select(treatment, PPTESTCD, PPORRES) |>
  pivot_wider(names_from = PPTESTCD, values_from = PPORRES)

knitr::kable(nca_tbl,
             caption = "Simulated trontinemab plasma NCA per dose group (nM, nM*h, h)")
Simulated trontinemab plasma NCA per dose group (nM, nM*h, h)
treatment auclast cmax tmax tlast clast.obs lambda.z r.squared adj.r.squared lambda.z.time.first lambda.z.time.last lambda.z.n.points clast.pred half.life span.ratio aucinf.obs
0.1 mg/kg 863.800 10.98751 0.5 168 2.348425 0.0072585 0.9999107 0.9999068 134.2185 168 25 2.345282 95.49467 0.3537528 1187.342
0.4 mg/kg 3439.817 43.91784 0.5 168 9.269410 0.0073760 0.9999059 0.9999020 132.8109 168 26 9.255869 93.97393 0.3744557 4696.524
1.2 mg/kg 10207.035 131.48215 0.5 168 26.983444 0.0076107 0.9999057 0.9999019 131.4034 168 27 26.940816 91.07506 0.4018294 13752.486
3.6 mg/kg 29752.623 391.59372 0.5 168 75.808999 0.0080382 0.9999071 0.9999037 128.5882 168 29 75.672376 86.23195 0.4570437 39183.748
7.2 mg/kg 57218.906 771.68380 0.5 168 141.491969 0.0083788 0.9999043 0.9999011 124.3655 168 32 141.190457 82.72621 0.5274562 74105.788

Assumptions and deviations

  • Brain homogenate equation simplification. The supplement equation for CBrHomo0 includes a contribution from ventricular CSF A(12) * (VCSFLV + VCSFTFV) and a denominator (VBtotal + VCSFLV + VCSFTFV). The volumes VBtotal, VCSFLV, and VCSFTFV are Bloomingdale 2017 model constants that are not reported in the Muliaditan 2025 paper, supplement, or any other on-disk source for this extraction. The model file therefore approximates Cbrain as a volume-weighted average over the (BBB endosomal + brain ISF + BCSFB endosomal) compartments only, scaled by the estimated FAC_BPRED = 0.05 correction. This is a strict subset of the published equation and may under-predict brain homogenate by a small amount that depends on ventricular CSF contribution. Users who want the full equation can re-derive VBtotal, VCSFLV, and VCSFTFV from the Bloomingdale 2017 source publication and pass them through a wrapper that recomputes Cbrain post hoc from the rxode2 state vector.
  • Mixture model selection. The paper’s NONMEM mixture model assigns each subject to either POP1 (fast kint, fraction 0.437) or POP2 (slow kint, fraction 0.563) without correlation to KD,TfR or data source. The model file exposes this as the canonical binary covariate MIX_FAST_ELIM. Typical-value simulation uses MIX_FAST_ELIM = 0 (slow, dominant subpopulation); population simulations should draw MIX_FAST_ELIM ~ Bernoulli(0.437) per subject.
  • No inter-individual variability. The paper did not estimate IIV because the NHP dataset was largely mean digitised profiles. The model files therefore carry no eta* parameters; rxode2::zeroRe() is cosmetic on these models since there is nothing to zero.
  • Per-compound TfR binding inputs. kon_T and koff_T are NOT population estimates - they are biophysical inputs taken from each compound’s binding assay (paper Table S2). The default ini() encodes a non-TfR control IgG (kon_T = 1e-12 placeholder). To simulate a specific antibody, override lkon_t, lkoff_t, and mw_kda via the params = argument to rxSolve().
  • Plasma observation is TOTAL drug. Per paper Methods, plasma measurements are total drug (free + TfR-bound complex). The model file observation Cc <- c_plasma + c_complex_plasma matches this; CSF and brain observations are unbound / homogenate concentrations respectively.
  • Allometric scaling between NHP and human. Both kint POP1 and kint POP2 are scaled from NHP to human by (70/6.2)^(-0.25) = 0.546 per the paper’s stated exponent. The paper explicitly reports the POP1 scaling (0.0329 -> 0.0179) but does not tabulate the POP2 scaling; the human file applies the same allometric formula to POP2 (0.0125 -> 0.00683).
  • Synthesis rates for luminal BBB / BCSFB TfR. Ksyn_uTfRBBB and Ksyn_uTfRBCSFB are not reported as separate parameters in the paper. At steady state without drug they are constrained by the baseline: Ksyn = Kdeg * uTFR0. The model file derives them inline from Kdeg_uTfR and uTFR0 per this steady-state identity.
  • FcRn baseline initial conditions. The supplement states “The initial condition for all the compartments is 0, except for unbound FcRn (A14-A16)”. The model file initialises the free-FcRn compartments at FcRn * V (where FcRn = 4.98e-5 mol/L from Supp Table S1).