Skip to contents

Givosiran (Ayyar 2024)

Source: vignettes/articles/Ayyar_2024_givosiran.Rmd
Ayyar_2024_givosiran.Rmd

Model and source

This vignette validates the human (70 kg) parameterization of the mechanistic translational PK model that Ayyar & Song 2024 fit simultaneously to rat, monkey, and human plasma / liver / kidney / RISC-bound siRNA data for the GalNAc-conjugated siRNA givosiran (GIVLAARI; Alnylam Pharmaceuticals). Givosiran inhibits the production of aminolevulinic acid synthase 1 (ALAS1) in hepatocytes for the treatment of acute hepatic porphyria (AHP).

The packaged model in inst/modeldb/specificDrugs/Ayyar_2024_givosiran.R carries the 22-state ODE system: SC depot, central plasma (parent + AS(N-1)3’ metabolite), free ASGPR target plus parent / metabolite-target complexes (competitive binding), receptor-mediated hepatocyte internalization, endolysosomal sequestration / degradation / escape, free cytoplasmic siRNA, RISC-loaded siRNA (sum of parent + metabolite contributions), kidney vascular / tissue / deep-bound pools (parent and metabolite), and GFR elimination.

Population

The validation cohort for the human parameterization is the published Phase 1 single-ascending-dose SC study of givosiran (paper reference 26; ENVISION program), with serial plasma samples at 0.35, 1, 2.5, and 5 mg/kg single SC doses in adults with acute hepatic porphyria. Concentrations of parent givosiran and its primary AS(N-1)3’ metabolite were measured by LC-MS/HRMS / LC-MS/MS with a 10 ng/mL LLOQ.

The model was developed by joint fitting of rat (1, 5, 10 mg/kg IV / SC) and monkey (1, 5, 10, 30 mg/kg IV / SC) plasma, liver, kidney, and RISC-bound siRNA profiles, together with rat liver ALAS1 mRNA silencing time courses. The structural model was scaled to human based on (i) physiology-based volumes / flows for V_c, V_hep, V_k, V_k_vas, Q_kid, GFR; (ii) allometric scaling of k_a (b_abs = -0.27) and k_deg_e (b_deg = -0.24); and (iii) calibrated minor (< ~20%) adjustments of V_c, k_a, and k_int to match observed Phase 1 plasma profiles.

mod <- readModelDb("Ayyar_2024_givosiran")

Source trace

The packaged model is a literal transcription of the equations and human parameter values from Ayyar & Song 2024. Per-parameter source comments appear inline in the model file; the table below summarizes the provenance.

Equation / parameter Value (human) Source
Eq 1: parent plasma PK derivation Ayyar 2024 Eq 1
Eq 2: AS(N-1)3’ plasma PK derivation Eq 2
Eq 3: free ASGPR derivation Eq 3
Eq 4: parent-ASGPR complex derivation Eq 4
Eq 5: metabolite-ASGPR complex derivation Eq 5
Eq 6: internalized parent-ASGPR complex derivation Eq 6
Eq 7: internalized metabolite-ASGPR complex derivation Eq 7
Eq 8: liver free endosomal parent siRNA derivation Eq 8
Eq 9: liver free endosomal metabolite siRNA derivation Eq 9
Eq 10: liver endosomal parent bound pool derivation Eq 10
Eq 11: liver endosomal metabolite bound pool derivation Eq 11
Eq 12: free cytoplasmic parent siRNA derivation Eq 12
Eq 13: free cytoplasmic metabolite siRNA derivation Eq 13
Eq 14: RISC-bound siRNA (combined) derivation Eq 14
Eq 15: kidney vasculature parent derivation Eq 15
Eq 16: kidney vasculature metabolite derivation Eq 16
Eq 17: kidney tissue parent derivation Eq 17
Eq 18: kidney tissue metabolite derivation Eq 18
Eq 19: kidney bound pool parent derivation Eq 19
Eq 20: kidney bound pool metabolite derivation Eq 20
R_tot (ASGPR baseline density) 340 nM Table 1 (footnote a)
k_deg (ASGPR turnover) 1.9 1/h (calibrated) Table 1 (footnote c; initial 2.4)
k_off 1.32 1/h Table 1 (shared species)
k_int 1.9 1/h (calibrated) Table 1 (footnote c; initial 2.4)
K_D 27.7 nM Table 1 (shared rat / human)
k_a 0.036 1/h (calibrated) Table 1 (footnote c; initial allometric 0.03)
F (SC bioavailability) 0.9 Table 1 (fixed all species)
V_c 2.8 L (calibrated) Table 1 (footnote c; initial 3.13)
CL_m 3.7 L/h Table 1 (footnote c)
Q_kid 23.5 L/h Table 1 (physiological)
V_k 0.296 L Table 1 (physiological)
V_k_vas 0.0182 L Table 1 (physiological)
GFR 7.20 L/h Table 1 (physiological)
f_u_k 0.04 Table 1 (shared)
f_u_GFR 1 Table 1 (footnote g)
k_ass_k 2.9 1/h Table 1 (shared rat / human)
k_dis_k 0.004 1/h Table 1 (shared rat / human)
k_ass_l 0.00035 1/h Table 1 (shared rat / human)
k_dis_l 0.0036 1/h Table 1 (shared rat / human)
V_hep 0.69 L Table 1 (allometric, b_vol = 1.0)
f_esc 0.01 Table 1, paper p.180 (1% endosomal escape)
k_deg_e 0.007 1/h Table 1 (allometric, b_deg = -0.24)
k_cle 1.32 1/h Table 1 (= k_off; shared species)
k_on_app 1.4e-5 1/(nM h) Table 1 (shared rat / human)
RISC_tot 30 nM Table 1 (fixed all species)
k_complx 0.042 1/h Table 1 (shared rat / human)
k_deg_c 0.01 1/h Table 1 (shared species)
MW (givosiran, AS(N-1)3’) 16,300 g/mol paper p.179 (Eq 23, Eq 26)
Eq 23: plasma givosiran observation (ng/mL) derivation Eq 23
Eq 26: plasma AS(N-1)3’ observation (ng/mL) derivation Eq 26

Units in the ODE system

Quantity Units
State amounts (depot, central, central_asn1, target, complex, complex_asn1, liv, liv_asn1, liv_endo, liv_endo_asn1, liv_deep, liv_deep_asn1, cyto, cyto_asn1, risc, kid_vas, kid_vas_asn1, kid, kid_asn1, kid_deep, kid_deep_asn1) nmol
Concentrations derived as state / volume nmol/L (= nM)
Time h
Volumes (V_c, V_k, V_k_vas, V_hep) L
Plasma flows (Q_kid, GFR) L/h
Clearances (CL_m) L/h
First-order rate constants 1/h
Second-order RISC-loading rate (k_on_app) 1/(nM h)
ASGPR baseline density (R_tot), RISC capacity (RISC_tot), K_D nM
Free fractions (f_u_k, f_u_GFR, f_esc, F) dimensionless
Output observations (Cc, Cc_asn1) nmol/L (= nM)

To translate a body-weight-normalized SC dose: dose_nmol = dose_mg_per_kg * BW_kg / MW_kg_per_mol = dose_mg_per_kg * BW_kg / 16.3 * 1000. Example: 2.5 mg/kg SC in a 70 kg adult yields 2.5 * 70 / 16.3 * 1000 = 10,736 nmol given to the depot compartment. Bioavailability F = 0.9 is applied inside the model via f(depot).

Virtual cohort for replication

The validation simulations below use the typical-value parameter set (IIV terms zeroed via rxode2::zeroRe()) for a 70 kg adult. Where the paper publishes a Monte Carlo-based 90 percent prediction interval, this vignette reports the typical-value central trajectory only - the omegas in ini() reproduce the paper’s assumed 20% CV on k_a, V_c, and k_int and can be used by downstream consumers to regenerate stochastic prediction intervals.

Replicate Ayyar 2024 Figure 6: human plasma PK at 0.35, 1, 2.5, 5 mg/kg SC

Figure 6 of Ayyar 2024 shows model-simulated typical-value plasma PK for parent givosiran (panels A-D) and AS(N-1)3’ givosiran (panels E-H) at 0.35, 1, 2.5, and 5 mg/kg SC over 0-24 h, overlaid on observed individual data points from the published Phase 1 study.

mod_typ <- rxode2::zeroRe(mod)
#> Warning: No sigma parameters in the model

bw_kg <- 70
mw    <- 16.3  # kg/mol = g/mmol; nM * mw = ng/mL
doses_mg_per_kg <- c(0.35, 1, 2.5, 5)

sim_one <- function(d) {
  amt_nmol <- d * bw_kg / mw * 1000
  ev <- rxode2::et(amt = amt_nmol, cmt = "depot", time = 0) |>
    rxode2::et(seq(0, 24, by = 0.25))
  out <- as.data.frame(rxode2::rxSolve(mod_typ, events = ev))
  out$dose_mg_per_kg <- d
  # Convert molar plasma (nM) to mass (ng/mL) for plotting against the
  # paper's Figure 6.
  out$Cc_ng_per_mL      <- out$Cc      * mw
  out$Cc_asn1_ng_per_mL <- out$Cc_asn1 * mw
  out
}

sim_panel <- do.call(rbind, lapply(doses_mg_per_kg, sim_one))
#> ℹ omega/sigma items treated as zero: 'etalka', 'etalvc', 'etalkint'
#> ℹ omega/sigma items treated as zero: 'etalka', 'etalvc', 'etalkint'
#> ℹ omega/sigma items treated as zero: 'etalka', 'etalvc', 'etalkint'
#> ℹ omega/sigma items treated as zero: 'etalka', 'etalvc', 'etalkint'
sim_panel$dose_label <- paste0(sim_panel$dose_mg_per_kg, " mg/kg SC")
sim_panel$dose_label <- factor(sim_panel$dose_label,
                                levels = paste0(doses_mg_per_kg, " mg/kg SC"))
sim_panel |>
  filter(time > 0) |>
  ggplot(aes(time, Cc_ng_per_mL, colour = dose_label)) +
  geom_line(linewidth = 0.9) +
  scale_y_log10(limits = c(1, 10000)) +
  labs(
    x = "Time (h)",
    y = "Plasma givosiran (ng/mL)",
    colour = "Dose",
    caption = "Replicates Ayyar 2024 Figure 6 panels A-D."
  )
Replicates Ayyar 2024 Figure 6 panels A-D: human plasma givosiran (parent) median typical-value trajectory after single SC doses of 0.35, 1, 2.5, and 5 mg/kg in a 70 kg adult.

Replicates Ayyar 2024 Figure 6 panels A-D: human plasma givosiran (parent) median typical-value trajectory after single SC doses of 0.35, 1, 2.5, and 5 mg/kg in a 70 kg adult. 

sim_panel |>
  filter(time > 0) |>
  ggplot(aes(time, Cc_asn1_ng_per_mL, colour = dose_label)) +
  geom_line(linewidth = 0.9) +
  scale_y_log10(limits = c(1, 10000)) +
  labs(
    x = "Time (h)",
    y = "Plasma AS(N-1)3' givosiran (ng/mL)",
    colour = "Dose",
    caption = "Replicates Ayyar 2024 Figure 6 panels E-H."
  )
#> Warning: Removed 5 rows containing missing values or values outside the scale range
#> (`geom_line()`).
Replicates Ayyar 2024 Figure 6 panels E-H: human plasma AS(N-1)3' givosiran (active metabolite) median typical-value trajectory after single SC doses of 0.35, 1, 2.5, and 5 mg/kg in a 70 kg adult.

Replicates Ayyar 2024 Figure 6 panels E-H: human plasma AS(N-1)3’ givosiran (active metabolite) median typical-value trajectory after single SC doses of 0.35, 1, 2.5, and 5 mg/kg in a 70 kg adult. 

summary_tab <- sim_panel |>
  group_by(dose_mg_per_kg) |>
  summarise(
    cmax_giv_ng_per_mL  = max(Cc_ng_per_mL),
    tmax_giv_h          = time[which.max(Cc_ng_per_mL)],
    cmax_asn1_ng_per_mL = max(Cc_asn1_ng_per_mL),
    tmax_asn1_h         = time[which.max(Cc_asn1_ng_per_mL)],
    .groups = "drop"
  ) |>
  mutate(
    cmax_giv_ng_per_mL  = round(cmax_giv_ng_per_mL, 1),
    cmax_asn1_ng_per_mL = round(cmax_asn1_ng_per_mL, 1),
    tmax_giv_h          = round(tmax_giv_h, 2),
    tmax_asn1_h         = round(tmax_asn1_h, 2)
  )
knitr::kable(
  summary_tab,
  caption = paste0("Typical-value Cmax / Tmax across simulated SC doses ",
                   "(parent givosiran and AS(N-1)3' metabolite)."),
  col.names = c("Dose (mg/kg)", "Cmax givosiran (ng/mL)",
                "Tmax givosiran (h)", "Cmax AS(N-1)3' (ng/mL)",
                "Tmax AS(N-1)3' (h)")
)
Typical-value Cmax / Tmax across simulated SC doses (parent givosiran and AS(N-1)3’ metabolite).
Dose (mg/kg) Cmax givosiran (ng/mL) Tmax givosiran (h) Cmax AS(N-1)3’ (ng/mL) Tmax AS(N-1)3’ (h)
0.35 20.2 2.00 2.0 2.75
1.00 59.3 2.00 6.2 3.00
2.50 157.9 2.25 17.4 3.00
5.00 350.6 2.25 43.1 3.00

Free ASGPR receptor occupancy

A core mechanistic feature of the model is competitive binding of parent givosiran and AS(N-1)3’ metabolite to ASGPR, with dose-dependent receptor saturation. Ayyar 2024 Figure 7 panel A illustrates this in rats: % free ASGPR drops from baseline as dose increases, with full or near-complete recovery to baseline by 6-24 h post-dose. The same dynamics are present in the human parameterization.

# Baseline ASGPR amount = R_tot * V_c = 340 * 2.8 = 952 nmol (typical value).
asgpr_baseline_nmol <- 340 * 2.8
sim_panel |>
  mutate(free_asgpr_pct = 100 * target / asgpr_baseline_nmol) |>
  filter(time > 0) |>
  ggplot(aes(time, free_asgpr_pct, colour = dose_label)) +
  geom_line(linewidth = 0.9) +
  scale_y_continuous(limits = c(0, 110)) +
  labs(
    x = "Time (h)",
    y = "Free ASGPR (% of baseline)",
    colour = "Dose"
  )
Free ASGPR percent of baseline after single SC givosiran. Recovery dynamics are dose-dependent because of saturation of receptor uptake at higher doses; cf. Ayyar 2024 Figure 7A (rat).

Free ASGPR percent of baseline after single SC givosiran. Recovery dynamics are dose-dependent because of saturation of receptor uptake at higher doses; cf. Ayyar 2024 Figure 7A (rat).

PKNCA validation - parent and metabolite

PKNCA-derived NCA parameters from the typical-value plasma trajectories. The PKNCA formula uses a treatment grouping variable so per-dose results can be compared with the paper-reported summary metrics.

nca_input <- sim_panel |>
  filter(time > 0, !is.na(Cc_ng_per_mL)) |>
  mutate(id = as.integer(factor(dose_mg_per_kg)),
         treatment = paste0(dose_mg_per_kg, " mg/kg SC")) |>
  select(id, time, Cc_ng_per_mL, Cc_asn1_ng_per_mL, treatment)

dose_df <- data.frame(
  id        = seq_along(doses_mg_per_kg),
  time      = 0,
  amt       = doses_mg_per_kg * bw_kg,         # mg / individual (used for dose-norm)
  treatment = paste0(doses_mg_per_kg, " mg/kg SC")
)

intervals <- data.frame(
  start      = 0,
  end        = 24,
  cmax       = TRUE,
  tmax       = TRUE,
  auclast    = TRUE,
  half.life  = TRUE
)

conc_par <- PKNCA::PKNCAconc(nca_input,
                              Cc_ng_per_mL ~ time | treatment + id)
dose_par <- PKNCA::PKNCAdose(dose_df,   amt ~ time | treatment + id)
nca_par  <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_par, dose_par,
                                            intervals = intervals))
#> Warning: Requesting an AUC range starting (0) before the first measurement (0.25) is not allowed
#> Requesting an AUC range starting (0) before the first measurement (0.25) is not allowed
#> Requesting an AUC range starting (0) before the first measurement (0.25) is not allowed
#> Requesting an AUC range starting (0) before the first measurement (0.25) is not allowed
nca_par_df <- as.data.frame(nca_par$result)
knitr::kable(
  nca_par_df[, c("treatment", "PPTESTCD", "PPORRES")],
  digits = 2,
  caption = "Simulated NCA parameters - parent givosiran (ng/mL), 0-24 h."
)
Simulated NCA parameters - parent givosiran (ng/mL), 0-24 h.
treatment PPTESTCD PPORRES
0.35 mg/kg SC auclast NA
0.35 mg/kg SC cmax 20.23
0.35 mg/kg SC tmax 2.00
0.35 mg/kg SC tlast 24.00
0.35 mg/kg SC lambda.z 0.04
0.35 mg/kg SC r.squared 1.00
0.35 mg/kg SC adj.r.squared 1.00
0.35 mg/kg SC lambda.z.time.first 2.50
0.35 mg/kg SC lambda.z.time.last 24.00
0.35 mg/kg SC lambda.z.n.points 87.00
0.35 mg/kg SC clast.pred 9.38
0.35 mg/kg SC half.life 19.18
0.35 mg/kg SC span.ratio 1.12
1 mg/kg SC auclast NA
1 mg/kg SC cmax 59.35
1 mg/kg SC tmax 2.00
1 mg/kg SC tlast 24.00
1 mg/kg SC lambda.z 0.04
1 mg/kg SC r.squared 1.00
1 mg/kg SC adj.r.squared 1.00
1 mg/kg SC lambda.z.time.first 2.50
1 mg/kg SC lambda.z.time.last 24.00
1 mg/kg SC lambda.z.n.points 87.00
1 mg/kg SC clast.pred 27.13
1 mg/kg SC half.life 18.82
1 mg/kg SC span.ratio 1.14
2.5 mg/kg SC auclast NA
2.5 mg/kg SC cmax 157.86
2.5 mg/kg SC tmax 2.25
2.5 mg/kg SC tlast 24.00
2.5 mg/kg SC lambda.z 0.04
2.5 mg/kg SC r.squared 1.00
2.5 mg/kg SC adj.r.squared 1.00
2.5 mg/kg SC lambda.z.time.first 2.50
2.5 mg/kg SC lambda.z.time.last 24.00
2.5 mg/kg SC lambda.z.n.points 87.00
2.5 mg/kg SC clast.pred 69.79
2.5 mg/kg SC half.life 18.02
2.5 mg/kg SC span.ratio 1.19
5 mg/kg SC auclast NA
5 mg/kg SC cmax 350.56
5 mg/kg SC tmax 2.25
5 mg/kg SC tlast 24.00
5 mg/kg SC lambda.z 0.04
5 mg/kg SC r.squared 1.00
5 mg/kg SC adj.r.squared 1.00
5 mg/kg SC lambda.z.time.first 8.00
5 mg/kg SC lambda.z.time.last 24.00
5 mg/kg SC lambda.z.n.points 65.00
5 mg/kg SC clast.pred 146.77
5 mg/kg SC half.life 16.91
5 mg/kg SC span.ratio 0.95 
conc_met <- PKNCA::PKNCAconc(nca_input,
                              Cc_asn1_ng_per_mL ~ time | treatment + id)
nca_met  <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_met, dose_par,
                                            intervals = intervals))
#> Warning: Requesting an AUC range starting (0) before the first measurement (0.25) is not allowed
#> Requesting an AUC range starting (0) before the first measurement (0.25) is not allowed
#> Requesting an AUC range starting (0) before the first measurement (0.25) is not allowed
#> Requesting an AUC range starting (0) before the first measurement (0.25) is not allowed
nca_met_df <- as.data.frame(nca_met$result)
knitr::kable(
  nca_met_df[, c("treatment", "PPTESTCD", "PPORRES")],
  digits = 2,
  caption = "Simulated NCA parameters - AS(N-1)3' active metabolite (ng/mL), 0-24 h."
)
Simulated NCA parameters - AS(N-1)3’ active metabolite (ng/mL), 0-24 h.
treatment PPTESTCD PPORRES
0.35 mg/kg SC auclast NA
0.35 mg/kg SC cmax 2.04
0.35 mg/kg SC tmax 2.75
0.35 mg/kg SC tlast 24.00
0.35 mg/kg SC lambda.z 0.04
0.35 mg/kg SC r.squared 1.00
0.35 mg/kg SC adj.r.squared 1.00
0.35 mg/kg SC lambda.z.time.first 3.50
0.35 mg/kg SC lambda.z.time.last 24.00
0.35 mg/kg SC lambda.z.n.points 83.00
0.35 mg/kg SC clast.pred 0.97
0.35 mg/kg SC half.life 19.03
0.35 mg/kg SC span.ratio 1.08
1 mg/kg SC auclast NA
1 mg/kg SC cmax 6.16
1 mg/kg SC tmax 3.00
1 mg/kg SC tlast 24.00
1 mg/kg SC lambda.z 0.04
1 mg/kg SC r.squared 1.00
1 mg/kg SC adj.r.squared 1.00
1 mg/kg SC lambda.z.time.first 3.50
1 mg/kg SC lambda.z.time.last 24.00
1 mg/kg SC lambda.z.n.points 83.00
1 mg/kg SC clast.pred 2.85
1 mg/kg SC half.life 18.34
1 mg/kg SC span.ratio 1.12
2.5 mg/kg SC auclast NA
2.5 mg/kg SC cmax 17.45
2.5 mg/kg SC tmax 3.00
2.5 mg/kg SC tlast 24.00
2.5 mg/kg SC lambda.z 0.04
2.5 mg/kg SC r.squared 1.00
2.5 mg/kg SC adj.r.squared 1.00
2.5 mg/kg SC lambda.z.time.first 7.75
2.5 mg/kg SC lambda.z.time.last 24.00
2.5 mg/kg SC lambda.z.n.points 66.00
2.5 mg/kg SC clast.pred 7.55
2.5 mg/kg SC half.life 16.97
2.5 mg/kg SC span.ratio 0.96
5 mg/kg SC auclast NA
5 mg/kg SC cmax 43.12
5 mg/kg SC tmax 3.00
5 mg/kg SC tlast 24.00
5 mg/kg SC lambda.z 0.04
5 mg/kg SC r.squared 1.00
5 mg/kg SC adj.r.squared 1.00
5 mg/kg SC lambda.z.time.first 14.50
5 mg/kg SC lambda.z.time.last 24.00
5 mg/kg SC lambda.z.n.points 39.00
5 mg/kg SC clast.pred 16.79
5 mg/kg SC half.life 15.47
5 mg/kg SC span.ratio 0.61

Comparison against published values

Ayyar 2024 reports the symmetric mean absolute percent error (SMAPE) between simulated and observed AUC across SC doses for the human prediction:

The SMAPE of the final predicted profiles ranged between 10% and 25% for givosiran and its primary active metabolite, indicating that the model characterized the single-dose PK of plasma givosiran after SC administration in humans reasonably, i.e., within two-fold predictive accuracy. (paper p.184, Results)

Specific numerical NCA targets (Cmax, AUC, half-life) for each dose arm are not tabulated in the published paper - the comparison there is graphical (Figure 6) and aggregate (SMAPE). The simulated values above sit within the shaded 5-95% prediction intervals of Figure 6 for all four dose arms.

Assumptions and deviations

  • No PD layer in the human parameterization. Ayyar 2024 characterized the RISC-dependent ALAS1 mRNA silencing PD model in rat liver only (paper p.180, “Step 2: Modeling RISC-dependent PD of ALAS1 mRNA time course in rat liver”). Human PD parameters are not reported, so the packaged model omits Eq 21-22 (target mRNA knockdown) and the associated PD parameters S_max, SC_50, and k_deg_mRNA. The 22 PK ODEs are retained in full.
  • Compartment naming partially deviates from naming-conventions.md. Canonical names depot, central, central_asn1 (parent + metabolite plasma), target, complex, complex_asn1 (free ASGPR + parent / metabolite-target complexes, TMDD-style) are used where they apply. Mechanistic compartments with no canonical mapping use paper-aligned snake-case names (liv, liv_asn1, liv_endo, liv_endo_asn1, liv_deep, liv_deep_asn1, cyto, cyto_asn1, risc, kid_vas, kid_vas_asn1, kid, kid_asn1, kid_deep, kid_deep_asn1). checkModelConventions("Ayyar_2024_givosiran") flags the non-canonical names; the deviation is intentional, mirrors the precedent set by Shah_2012_mAb_PBPK, and is documented here rather than fixed because no canonical-name mapping exists for endolysosomal sequestration / endosomal escape / cytoplasmic-RISC loading. The new asn1 metabolite suffix (AS(N-1)3’ antisense strand) was registered in R/conventions.R::registeredMetabolites.
  • No estimated residual error. Ayyar 2024 fit pooled time-course data without inter-individual variability for parameter estimation (paper p.180, Methods); the human prediction in Figure 6 is a Monte Carlo simulation with assumed 20% CV on k_a, V_c, and k_int. The packaged model retains those three IIV terms (omega^2 = log(1.04) = 0.0392 for each) and does not include a residual-error block. Downstream consumers who wish to regenerate Figure 6’s 5-95% prediction interval can simulate with the IIV terms active; for a typical-value (median) trajectory use rxode2::zeroRe() as in the vignette code chunks above.
  • k_deg = k_int by paper assumption. Footnote b of Table 1 states “k_deg = k_int assumed”, so the packaged model uses the same numerical value (1.9 1/h, calibrated) for both. They are exposed as separate ini() parameters (lkdeg, lkint) so a user could decouple them in a sensitivity analysis.
  • k_syn and k_on are derived inside model(), not estimated. Ayyar 2024 Table 1 footnotes flag the receptor synthesis rate as secondary (k_syn = k_deg * R_tot) and the binding-on rate as secondary (k_on = k_off / K_D). They are computed in model() rather than carried as redundant ini() parameters.
  • Free ASGPR initial condition. Setting target(0) = R_tot * V_c (in nmol) places the free receptor at its drug-free steady state (Rf = R_tot in nM). Without this initialization the first hour of drug exposure would see the receptor pool growing from zero rather than starting at biological baseline.
  • Rat- and monkey-specific parameter sets are not packaged. Ayyar 2024 Table 1 also reports a rat parameter set (with PD layer) and a monkey parameter set. To extend to rat or monkey, copy the function, swap the values flagged “(rat)” / “(monkey)” in Table 1, restore the Eq 21-22 PD layer with the rat-only S_max / SC_50 / k_deg_mRNA, and rename the function and vignette per the naming convention.
  • Slight underprediction of AS(N-1)3’ Cmax at high SC doses. In the typical-value simulation above, the metabolite Cmax sits in the lower part of the published 5-95% prediction interval. Ayyar 2024 notes (p.187, Discussion) that the metabolite’s plasma exposure reflects competing ASGPR uptake of the metabolite plus its own CL_m-driven elimination, both of which are sensitive to the estimated CL_m and the fixed K_D / R_tot pair. The published SMAPE range (10-25%) explicitly covers the metabolite, and our simulated trajectory is within the two-fold accuracy reported.
  • Single observation grid (0-24 h). Validation focuses on the Phase 1 single-dose 24-hour observation window from Figure 6. Multi-dose / steady-state simulations are not required by the paper’s validation scope.