library(shiny)
library(bslib)
library(mrgsolve)
library(dplyr)
library(ggplot2)

# Tramadol PK/PD model: 2-compartment parent + 1-compartment active M1 metabolite + Imax PD
# Based on Chun et al. 2025, CPT: Pharmacometrics & Systems Pharmacology
# Racemic tramadol PK with (1R,2R)-M1 active metabolite linked to pain response
model_code <- '
$PARAM @annotated
CL   : 35     : Apparent clearance CL/F (L/h)
VC   : 200    : Central volume Vc/F (L)
VP   : 80     : Peripheral volume Vp/F (L)
Q    : 30     : Intercompartmental clearance Q/F (L/h)
KA   : 1.2    : Absorption rate constant (1/h)
TLAG : 0.30   : Absorption lag time (h)
FM1  : 0.269  : Fraction of racemic dose to active (1R,2R)-M1
CLM1 : 10     : M1 apparent clearance (L/h)
VM1  : 31     : M1 apparent volume (L)
KE0  : 0.0398 : Effect compartment rate constant (1/h)
IMAX : 0.908  : Maximum fractional pain reduction
PD0  : 46.4   : Baseline VAS pain score (mm)
IC50U : 0.588 : IC50 unbound (1R,2R)-M1 (ng/mL)
FUM1 : 0.246  : Fraction unbound of M1
WT   : 72     : Body weight (kg)
BMI  : 26     : Body mass index (kg/m2)
CYP2D6 : 1.0  : CYP2D6 activity multiplier

$CMT @annotated
DEPOT  : Oral depot (mg)
CENT   : Central tramadol (mg)
PERIPH : Peripheral tramadol (mg)
M1CENT : Central M1 (mg)
EFFECT : Effect compartment (conc scale)

$MAIN
ALAG_DEPOT = TLAG;
double CLi  = CL * pow(WT / 72.0, 0.75);
double VCi  = VC * pow(BMI / 26.0, 1.5);
double FM1i = FM1 * pow(CYP2D6, 0.5);

$ODE
double CM1_mgL = M1CENT / VM1;
dxdt_DEPOT  = -KA * DEPOT;
dxdt_CENT   = KA * DEPOT - (CLi / VCi) * CENT - (Q / VCi) * CENT + (Q / VP) * PERIPH;
dxdt_PERIPH = (Q / VCi) * CENT - (Q / VP) * PERIPH;
dxdt_M1CENT = FM1i * (CLi / VCi) * CENT - (CLM1 / VM1) * M1CENT;
dxdt_EFFECT = KE0 * (CM1_mgL - EFFECT);

$TABLE
double CP   = CENT / VCi * 1000.0;
double CM1  = M1CENT / VM1 * 1000.0;
double CE   = EFFECT * 1000.0;
double CUE  = FUM1 * CE;
double PAIN = PD0 * (1.0 - IMAX * CUE / (IC50U + CUE));

$CAPTURE CP CM1 PAIN
'

mod <- mcode("tramadol_pkpd", model_code)

app_theme <- bs_theme(
  version = 5, bootswatch = "flatly", primary = "#8b5cf6"
) |> bs_add_rules("
  .metric-card { background: #f8f9fa; border-radius: 8px; padding: 15px; margin: 5px; text-align: center; border: 1px solid #dee2e6; }
  .metric-value { font-size: 24px; font-weight: bold; color: #2c3e50; }
  .metric-label { font-size: 12px; color: #7f8c8d; }
  .metric-success .metric-value { color: #10b981; }
  .metric-warning .metric-value { color: #f59e0b; }
  .metric-primary .metric-value { color: #8b5cf6; }
  .metric-info .metric-value { color: #0dcaf0; }
  .ref-box { background: #f0f4ff; border-left: 4px solid #8b5cf6; padding: 12px 16px; border-radius: 4px; margin-top: 10px; font-size: 13px; }
  .ref-box a { color: #8b5cf6; }
")

ui <- page_sidebar(
  title = "Tramadol PK/PD Simulator \u2014 Neuropathic Pain",
  theme = app_theme,
  sidebar = sidebar(
    title = "Simulation Settings", width = 340,
    h6("Dosing"),
    sliderInput("dose", "Dose (mg)", min = 50, max = 400, value = 100, step = 50),
    selectInput("frequency", "Dosing Frequency",
                choices = c("Once daily (q24h)" = 24,
                            "Twice daily (q12h)" = 12,
                            "Three times daily (q8h)" = 8,
                            "Four times daily (q6h)" = 6),
                selected = 12),
    numericInput("n_days", "Duration (days)", value = 3, min = 1, max = 7),
    hr(), h6("Patient Characteristics"),
    sliderInput("wt", "Weight (kg)", min = 40, max = 120, value = 72, step = 1),
    sliderInput("bmi", "BMI (kg/m\u00B2)", min = 17, max = 40, value = 26, step = 1),
    selectInput("diabetes", "Diabetes Status",
                choices = c("None (nerve compression)" = "none",
                            "Type 1 Diabetes (T1DM)" = "t1dm",
                            "Type 2 Diabetes (T2DM)" = "t2dm"),
                selected = "none"),
    selectInput("cyp2d6", "CYP2D6 Phenotype",
                choices = c("Poor Metabolizer (PM)" = 0.2,
                            "Intermediate Metabolizer (IM)" = 0.5,
                            "Normal Metabolizer (NM/EM)" = 1.0,
                            "Ultra-rapid Metabolizer (UM)" = 2.0),
                selected = 1.0),
    hr(),
    checkboxInput("log_scale", "Log Scale (Y-axis)", value = FALSE),
    checkboxInput("show_therapeutic", "Show Therapeutic Range", value = TRUE)
  ),
  navset_card_tab(
    title = "Tramadol PK/PD Simulator",
    full_screen = TRUE,
    nav_panel("Simulation",
      layout_column_wrap(
        width = 1/4,
        fill = FALSE,
        div(class = "metric-card metric-success",
            div(class = "metric-value", textOutput("cmax")),
            div(class = "metric-label", "Cmax,ss (ng/mL)")),
        div(class = "metric-card metric-warning",
            div(class = "metric-value", textOutput("ctrough")),
            div(class = "metric-label", "Ctrough (ng/mL)")),
        div(class = "metric-card metric-primary",
            div(class = "metric-value", textOutput("auc")),
            div(class = "metric-label", "AUC\u2080-\u03C4 (ng\u00B7h/mL)")),
        div(class = "metric-card metric-info",
            div(class = "metric-value", textOutput("thalf")),
            div(class = "metric-label", "t\u00BD (h)"))
      ),
      plotOutput("pkPlot", height = "500px")
    ),
    nav_panel("Pain Response",
      layout_columns(
        col_widths = c(4, 4, 4),
        div(class = "metric-card metric-success",
            div(class = "metric-value", textOutput("pain_baseline")),
            div(class = "metric-label", "Baseline VAS (mm)")),
        div(class = "metric-card metric-primary",
            div(class = "metric-value", textOutput("pain_min")),
            div(class = "metric-label", "Min Pain Score (mm)")),
        div(class = "metric-card metric-warning",
            div(class = "metric-value", textOutput("pain_reduction")),
            div(class = "metric-label", "Max Pain Reduction (%)"))
      ),
      plotOutput("pdPlot", height = "500px")
    ),
    nav_panel("Model Information", markdown("
## Tramadol PK/PD Model \u2014 Neuropathic Pain with Diabetes Covariates

**Drug:** Tramadol (racemic mixture of (1R,2R)- and (1S,2S)-tramadol)
**Class:** Centrally acting opioid analgesic (\u00B5-opioid agonist + SNRI)
**Indication:** Neuropathic pain (diabetic neuropathy, nerve compression)
**Route:** Oral (IR tablets)

### PK Model Structure
- **Tramadol:** Two-compartment model with first-order absorption, lag time, and linear elimination
- **Active metabolite (1R,2R)-M1 (O-desmethyltramadol):** One-compartment with linear formation and elimination
- **(1R,2R)-M1** has 700\u00D7 higher \u00B5-opioid receptor affinity (Ki = 3.4 nM) than racemic tramadol

### PD Model Structure
- **Inhibitory Emax model** with effect compartment (ke0 = 0.0398 h\u207B\u00B9)
- PD linked to **unbound (1R,2R)-M1** concentration at effect site
- VAS pain score (0\u2013100 mm) as continuous endpoint

### Key Findings: Diabetes Effect
| Parameter | Non-Diabetic | T1DM | T2DM |
|-----------|-------------|------|------|
| fu (M1) | 0.246 | 0.756 | 0.284 |
| IC50 unbound (nmol/L) | 2.36 | 0.54 | 1.02 |
| Sensitivity | Reference | \u2191\u2191\u2191 | \u2191\u2191 |

- **Diabetic patients show enhanced tramadol sensitivity** (lower IC50)
- Higher fraction unbound in T1DM due to albumin glycation
- **Dosing guidance:** QD dosing sufficient for diabetic neuropathy; BID recommended for nerve compression neuropathy

### Covariates
- **BMI:** Power function on central volume of distribution
- **CYP2D6 activity:** Affects formation clearance of active M1 metabolite
- **Diabetes status:** Affects protein binding (fu) and pharmacodynamic sensitivity (IC50)
    ")),
    nav_panel("References", div(class = "ref-box",
      tags$h5("\U0001F4DA Key References"),
      tags$ol(
        tags$li("Chun D, Mehta P, Guzy S, et al. Enhanced Sensitivity to Tramadol in Diabetic Neuropathic Pain Compared to Nerve Compression Neuropathies: A Population PK/PD Model Analysis. CPT Pharmacometrics Syst Pharmacol. 2025;14(4):781-795. PMID: 39961992."),
        tags$li("Grond S, Sablotzki A. Clinical pharmacology of tramadol. Clin Pharmacokinet. 2004;43(13):879-923."),
        tags$li(tags$a(href = "https://www.accessdata.fda.gov/drugsatfda_docs/label/2009/020281s032s033lbl.pdf", target = "_blank", "ULTRAM (tramadol) FDA Prescribing Information")),
        tags$li("De Moraes NV, Lauretti GR, Lanchote VL. Effects of Type 1 and Type 2 Diabetes on the Pharmacokinetics of Tramadol Enantiomers. J Pharm Pharmacol. 2014;66(9):1222-1230.")
      ),
      tags$h5("\U0001F48A Therapeutic Context"),
      tags$ul(
        tags$li(tags$strong("Class:"), " Centrally acting opioid analgesic (\u00B5-opioid agonist + weak SNRI)"),
        tags$li(tags$strong("Active metabolite:"), " (1R,2R)-O-desmethyltramadol (M1) via CYP2D6"),
        tags$li(tags$strong("Dosing:"), " IR: 50\u2013100 mg q4\u20136h (max 400 mg/day)"),
        tags$li(tags$strong("Therapeutic range:"), " 100\u2013300 ng/mL (analgesic effect)"),
        tags$li(tags$strong("Half-life:"), " ~6 h (tramadol), ~7\u20139 h apparent (M1, formation-rate limited)"),
        tags$li(tags$strong("Metabolism:"), " CYP2D6 \u2192 M1 (active), CYP3A4/CYP2B6 \u2192 M2 (inactive)")
      ),
      tags$h5("\U0001F52C Clinical Implications"),
      tags$ul(
        tags$li("Diabetic neuropathic pain patients: \u226580% respond to 100 mg QD"),
        tags$li("Nerve compression neuropathy: requires BID dosing for sustained relief"),
        tags$li("CYP2D6 poor metabolizers have reduced M1 formation and diminished opioid effect"),
        tags$li("Tramadol has hypoglycemic effects in T2DM via \u00B5-opioid receptor activation")
      )
    ))
  ),
  div(style = "text-align: center; padding: 20px; margin-top: 30px; border-top: 1px solid #e9ecef; color: #6c757d; font-size: 12px;",
    "Powered by ", tags$a(href = "https://www.pkpdbuilder.com", target = "_blank",
      style = "color: #8b5cf6; font-weight: 500;", "PKPDBuilder.com"),
    " \u2022 Built by Sunny \u2600\uFE0F (Husain Attarwala's AI Assistant)", br(),
    tags$span(style = "font-size: 10px;", "For research and educational purposes only. Not for clinical decision-making."))
)

server <- function(input, output, session) {

  diabetes_params <- reactive({
    switch(input$diabetes,
      "none" = list(ic50u = 0.588, fum1 = 0.246, label = "Non-Diabetic"),
      "t1dm" = list(ic50u = 0.135, fum1 = 0.756, label = "T1DM"),
      "t2dm" = list(ic50u = 0.254, fum1 = 0.284, label = "T2DM")
    )
  })

  sim_data <- reactive({
    tau <- as.numeric(input$frequency)
    total_hours <- input$n_days * 24
    n_doses <- floor(total_hours / tau)
    dp <- diabetes_params()

    ev1 <- ev(amt = input$dose, cmt = 1, ii = tau, addl = max(n_doses - 1, 0))

    mod %>%
      param(WT = input$wt, BMI = input$bmi,
            CYP2D6 = as.numeric(input$cyp2d6),
            IC50U = dp$ic50u, FUM1 = dp$fum1) %>%
      ev(ev1) %>%
      mrgsim(end = total_hours, delta = 0.25) %>%
      as.data.frame() %>%
      mutate(
        time_h = time,
        time_days = time / 24
      )
  })

  last_interval_data <- reactive({
    d <- sim_data()
    tau <- as.numeric(input$frequency)
    total_hours <- input$n_days * 24
    last_dose_time <- total_hours - tau
    d %>% filter(time_h >= last_dose_time, time_h <= total_hours)
  })

  output$cmax <- renderText({
    li <- last_interval_data()
    if (nrow(li) < 2) return("--")
    sprintf("%.0f", max(li$CP, na.rm = TRUE))
  })

  output$ctrough <- renderText({
    li <- last_interval_data()
    if (nrow(li) < 2) return("--")
    sprintf("%.1f", min(li$CP, na.rm = TRUE))
  })

  output$auc <- renderText({
    li <- last_interval_data()
    if (nrow(li) < 2) return("--")
    auc_val <- sum(diff(li$time_h) * (head(li$CP, -1) + tail(li$CP, -1)) / 2)
    sprintf("%.0f", auc_val)
  })

  output$thalf <- renderText({
    CLi <- 35 * (input$wt / 72)^0.75
    VCi <- 200 * (input$bmi / 26)^1.5
    k10 <- CLi / VCi; k12 <- 30 / VCi; k21 <- 30 / 80
    s <- k10 + k12 + k21; p <- k10 * k21
    beta_val <- 0.5 * (s - sqrt(s^2 - 4 * p))
    sprintf("%.1f", log(2) / beta_val)
  })

  # Pain response metrics
  output$pain_baseline <- renderText({ sprintf("%.0f", 46.4) })

  output$pain_min <- renderText({
    d <- sim_data()
    sprintf("%.1f", min(d$PAIN, na.rm = TRUE))
  })

  output$pain_reduction <- renderText({
    d <- sim_data()
    min_pain <- min(d$PAIN, na.rm = TRUE)
    pct <- (46.4 - min_pain) / 46.4 * 100
    sprintf("%.0f%%", pct)
  })

  output$pkPlot <- renderPlot({
    d <- sim_data()
    dp <- diabetes_params()
    tau <- as.numeric(input$frequency)
    freq_label <- switch(as.character(tau), "24" = "QD", "12" = "BID", "8" = "TID", "6" = "QID")

    d_long <- d %>%
      tidyr::pivot_longer(cols = c(CP, CM1), names_to = "Analyte", values_to = "Conc") %>%
      mutate(Analyte = ifelse(Analyte == "CP", "Tramadol", "(1R,2R)-M1"))

    p <- ggplot(d_long, aes(x = time_h, y = Conc, color = Analyte)) +
      geom_line(linewidth = 0.8)

    if (input$show_therapeutic) {
      p <- p +
        annotate("rect", xmin = -Inf, xmax = Inf, ymin = 100, ymax = 300,
                 fill = "#10b981", alpha = 0.10) +
        geom_hline(yintercept = c(100, 300), linetype = "dashed", color = "#10b981", alpha = 0.5) +
        annotate("text", x = max(d$time_h) * 0.02, y = 310,
                 label = "Therapeutic: 100-300 ng/mL", hjust = 0, size = 3.5, color = "#10b981")
    }

    p <- p +
      scale_color_manual(values = c("Tramadol" = "#8b5cf6", "(1R,2R)-M1" = "#f59e0b")) +
      labs(
        x = "Time (hours)",
        y = "Concentration (ng/mL)",
        title = paste0("Tramadol ", input$dose, " mg ", freq_label,
                       " \u2014 ", dp$label,
                       " (CYP2D6: ", names(which(c("0.2" = "PM", "0.5" = "IM",
                         "1" = "NM", "2" = "UM") == input$cyp2d6)[1]), ")"),
        color = "Analyte"
      ) +
      theme_minimal(base_size = 14) +
      theme(plot.title = element_text(face = "bold"),
            legend.position = "top")

    if (input$log_scale) p <- p + scale_y_log10()
    p
  })

  output$pdPlot <- renderPlot({
    d <- sim_data()
    dp <- diabetes_params()
    tau <- as.numeric(input$frequency)
    freq_label <- switch(as.character(tau), "24" = "QD", "12" = "BID", "8" = "TID", "6" = "QID")

    p <- ggplot(d, aes(x = time_h, y = PAIN)) +
      geom_hline(yintercept = 46.4, linetype = "dashed", color = "#6c757d", alpha = 0.6) +
      annotate("text", x = max(d$time_h) * 0.02, y = 48,
               label = "Baseline (no drug)", hjust = 0, size = 3.5, color = "#6c757d") +
      geom_hline(yintercept = 46.4 * 0.5, linetype = "dotted", color = "#10b981", alpha = 0.6) +
      annotate("text", x = max(d$time_h) * 0.02, y = 46.4 * 0.5 + 1.5,
               label = "50% reduction target", hjust = 0, size = 3.5, color = "#10b981") +
      annotate("rect", xmin = -Inf, xmax = Inf, ymin = 0, ymax = 46.4 * 0.5,
               fill = "#10b981", alpha = 0.06) +
      geom_line(color = "#e74c3c", linewidth = 1.0) +
      scale_y_continuous(limits = c(0, 55)) +
      labs(
        x = "Time (hours)",
        y = "VAS Pain Score (mm)",
        title = paste0("Pain Response \u2014 Tramadol ", input$dose, " mg ", freq_label,
                       " (", dp$label, ")"),
        subtitle = paste0("IC50 (unbound M1) = ", dp$ic50u, " ng/mL | ",
                          "fu(M1) = ", dp$fum1,
                          " | Imax = 0.908")
      ) +
      theme_minimal(base_size = 14) +
      theme(plot.title = element_text(face = "bold"))

    p
  })
}

shinyApp(ui = ui, server = server)