Pancreatic_Analysis.Rmd

---
title: "Bruckner Oncology Resistant Cancer Survival Analysis"
author: "AJ Book"
output:
  word_document: default
  pdf_document: default
  html_document: default
editor_options:
  markdown:
    wrap: 72
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
# Clear the entire environment
rm(list = ls())

setwd("C:/Users/ajboo/BookAbraham/RProjects/MZBSurvivalAnalysis")

# Define the output directory path
output_dir <- "output/Pancreatic_Output"

# Create the output directory if it doesn't exist
if (!dir.exists(output_dir)) {
  dir.create(output_dir)
}

# Set the output directory for plots
knitr::opts_chunk$set(fig.path = paste0(output_dir, "/plot", "-"))


```


## Load Libraries

This section is reserved for libraries we will use throughout this RMD
file and any imported modules

```{python imports}
# Importing necessary libraries
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from lifelines import KaplanMeierFitter, CoxPHFitter

```

```{r libraries, echo=TRUE}
library(tidyverse)
library(survival)
library(survminer)
library(ggsci)
library(knitr)
library(ggsurvfit)
library(gt)
library(reticulate)
library(maxstat)
```


Import the data file


```{python load and convert data}
# Define the function to load and preprocess data
def load_and_convert_data(file_path, cancer_type):
    # Load data from CSV file
    df = pd.read_csv(file_path)
    
    # Subset data for the specified cancer type
    cancer_df = df[df['Cancer_Type'] == cancer_type].copy()  # Create a copy
    
    # Convert selected columns to categorical variables
    factors = ['Gender', 'Cancer_Type', 'Prior_Tx', 'Resistant', 'Cancer_Status', 'Risk_Group_ALAN']
    cancer_df[factors] = cancer_df[factors].astype('category')

        # Print message indicating successful loading
    print("Data for", cancer_type, "loaded successfully.")
    
    return cancer_df

# Load and preprocess data for Pancreatic Cancer
pancreatic_df = load_and_convert_data("data/Organized_Bruckner_Data.csv", "Pancreatic Cancer")

# Load and preprocess data for Ovarian Cancer
#ovarian_df = load_and_convert_data("data/Organized_Bruckner_Data.csv", "Ovarian Cancer")


```
Subset for Resistant Data

```{python resistant subset}

def subset_by_resistant(df, resistant_value):
    """
    Subset a DataFrame by the value of the 'Resistant' column.
    
    Parameters:
        df (pd.DataFrame): The DataFrame to subset.
        resistant_value (str): The value to subset by (e.g., 'Resistant').
    
    Returns:
        pd.DataFrame: The subsetted DataFrame.
    """
    subset_df = df[df['Resistant'] == resistant_value].copy()
    print(f"Subsetted DataFrame for Resistant='{resistant_value}' created successfully.")
    return subset_df

# Subset data frames for Resistant cases
#resistant_ovarian_df = subset_by_resistant(ovarian_df, "Resistant")
#resistant_pancreatic_df = subset_by_resistant(pancreatic_df, "Resistant")


```

```{python subset by cancer Status}
def subset_by_cancer_status(df, cancer_status):
    """
    Subset a DataFrame by the value of the 'Cancer_Status' column.
    
    Parameters:
        df (pd.DataFrame): The DataFrame to subset.
        cancer_status (str): The value to subset by (e.g., 'NPT- Pancreatic Cancer').
    
    Returns:
        pd.DataFrame: The subsetted DataFrame.
    """
    subset_df = df[df['Cancer_Status'] == cancer_status].copy()
    print(f"Subsetted DataFrame for Cancer_Status='{cancer_status}' created successfully.")
    return subset_df

# Subset data frames for NPT- Pancreatic Cancer and Resistant- Pancreatic Cancer
npt_pancreatic_df = subset_by_cancer_status(pancreatic_df, 'NPT- Pancreatic Cancer')
resistant_pancreatic_df = subset_by_cancer_status(pancreatic_df, 'Resistant- Pancreatic Cancer')

```
```{python recode ALAN}
# Recode Risk_Group_ALAN column
# Define the bins and labels
bins = [-1, 0, 2, 4]
labels = ['Low_Risk', 'Intermediate_Risk', 'High_Risk']

# Recode Risk_Group_ALAN column based on Prognostic_Score_ALAN
resistant_pancreatic_df['Risk_Group_ALAN'] = pd.cut(resistant_pancreatic_df['Prognostic_Score_ALAN'], bins=bins, labels=labels, include_lowest=False)
npt_pancreatic_df['Risk_Group_ALAN'] = pd.cut(npt_pancreatic_df['Prognostic_Score_ALAN'], bins=bins, labels=labels, include_lowest=False)

```

Examine the variables within your data 
```{python examine subset}
# Glimpse at the subsetted data frames
print("\nSubsetted NPT Pancreatic Cancer Data Frame:")
print(npt_pancreatic_df.head())
print("\nSubsetted Resistant Pancreatic Cancer Data Frame:")
print(resistant_pancreatic_df.head())

```


Determine the types of class each column contains as its datatype
```{python examine type}
# Check data types of columns in DataFrame
print(npt_pancreatic_df.dtypes)
print(resistant_pancreatic_df.dtypes)

```
## Numeric Summary

Step 1: calculate the numeric statistics of the pancreatic_df
data frame #Note:You can specify percentiles, quantiles and normality or you can give specific percentiles depending on what you are interested in looking at this specific usage is looking at the 33rd and 67th percentiles of the data

Step 2: Create histograms, boxplots and distribution curves to visualize the descriptive statistics of the numeric variables.

```{python numeric summary}
def calculate_numeric_statistics(data):
    # Select only numeric columns
    numeric_data = data.select_dtypes(include=np.number)
    
    # Calculate descriptive statistics
    descriptive_stats = numeric_data.describe().transpose()
    
    # Calculate interquartile range (IQR) and include quantiles (25th, 50th, and 75th percentiles)
    quantiles = numeric_data.quantile([0.25, 0.5, 0.75], axis=0).transpose()
    quantiles["IQR"] = quantiles[0.75] - quantiles[0.25]
    quantiles.columns = ["Q1", "Median", "Q3", "IQR"]
    
    # Calculate additional percentiles (33rd and 67th)
    custom_percentiles = np.percentile(numeric_data, [33, 67], axis=0)
    custom_percentiles_df = pd.DataFrame(custom_percentiles.T, columns=["33rd Percentile", "67th Percentile"], index=numeric_data.columns)
    
    # Combine all statistics
    stats_combined = pd.concat([descriptive_stats, quantiles, custom_percentiles_df], axis=1)
    
    return stats_combined

# Create summary statistics table for NPT pancreatic cancer
npt_pancreatic_stats = calculate_numeric_statistics(npt_pancreatic_df)

# Create summary statistics table for resistant pancreatic cancer
resistant_pancreatic_stats = calculate_numeric_statistics(resistant_pancreatic_df)

# Display the tables
print("Summary statistics for NPT Pancreatic cancer:")
print(npt_pancreatic_stats)
print("\nSummary statistics for resistant pancreatic cancer:")
print(resistant_pancreatic_stats)
```

```{r advanced numeric summary}

#Load Util functions
source("Utils.R")

# Generate the first table for Resistant Pancreatic Cancer
resistant_pancreatic_table <- calc_num_stats(py$resistant_pancreatic_df, selected_labels = c("Quantiles", "Percentiles"), percentiles = c(33, 67), title = "Numeric Statistics for Resistant Pancreatic Cancer")

# Save the table as an image
gtsave(resistant_pancreatic_table, filename = file.path(output_dir, "resistant_pancreatic_table.png"))

# Generate the second table for NPT Pancreatic Cancer
npt_pancreatic_table <- calc_num_stats(py$npt_pancreatic_df, selected_labels = c("Quantiles", "Percentiles"), percentiles = c(33, 67), title = "Numeric Statistics for NPT Pancreatic Cancer")

# Save the table as an image
gtsave(npt_pancreatic_table, filename = file.path(output_dir, "npt_pancreatic_table.png"))


```

```{python numeric distribution, echo=FALSE}
import matplotlib.pyplot as plt
import seaborn as sns

# Define the function to plot histograms and boxplots for one variable
def plot_numeric_statistics(df, variable, subset):
    # Create subplots
    fig, (ax_box, ax_hist) = plt.subplots(2, sharex=True, gridspec_kw={"height_ratios": (.15, .85)})
    
    # Plot boxplot
    sns.boxplot(x=df[variable], ax=ax_box, color='orange', width=0.3, linewidth=1.5, showmeans=True, meanline=True,
                meanprops=dict(color='black', linestyle='--', linewidth=2),
                medianprops=dict(color='black', linewidth=2))
    ax_box.set_ylabel(variable)
    
    # Calculate mean and std_dev
    mean = df[variable].mean()
    std_dev = df[variable].std()
    
    # Plot histogram with density function
    sns.histplot(df[variable], kde=True, bins=12, stat='density', color='skyblue', ax=ax_hist)
    ax_hist.set_xlabel(variable)
    ax_hist.set_ylabel('Density')
    
    # Add lines for mean and mean +/- std_dev to the histogram
    ax_hist.axvline(mean, color='red', linestyle='--', linewidth=2, label=f'Mean: {mean:.2f}')
    ax_hist.axvline(mean + std_dev, color='purple', linestyle='--', linewidth=2, label=f'Mean + Std Dev: {mean + std_dev:.2f}')
    ax_hist.axvline(mean - std_dev, color='purple', linestyle='--', linewidth=2, label=f'Mean - Std Dev: {mean - std_dev:.2f}')
    
    # Add label for the IQR on the boxplot
    q1 = df[variable].quantile(0.25)
    q3 = df[variable].quantile(0.75)
    iqr = q3 - q1
    ax_box.text(0.5, 0.5, 'IQR', color='black', ha='center', fontsize=10, transform=ax_box.transAxes)
    
    # Remove y-axis ticks for boxplot
    ax_box.set_yticks([])
    
    # Despine the plots
    sns.despine(ax=ax_hist)
    sns.despine(ax=ax_box, left=True)
    
    # Set common xlabel
    plt.xlabel(variable)
    
    # Add title to the entire plot
    plt.suptitle(f'{subset} - by {variable}')
    
    # Show the plot
    plt.tight_layout()
    
    # Save the plot as an image
    plt.savefig(f'output/Pancreatic_Output/{subset}_{variable}_plot.png')
    
    # Close the plot to release memory
    plt.close()

# List of columns to exclude from numeric variables
exclude_columns = ['Prognostic_Score_ALAN', 'Event_Status']

# Iterate over each numeric variable in your dataset and call the plot_numeric_statistics function
for column in npt_pancreatic_df.select_dtypes(include=['int64', 'float64']).columns:
    if column not in exclude_columns:
        plot_numeric_statistics(npt_pancreatic_df, column, 'NPT_Pancreatic_Cancer')
        
# Iterate over each numeric variable in your dataset and call the plot_numeric_statistics function
for column in resistant_pancreatic_df.select_dtypes(include=['int64', 'float64']).columns:
    if column not in exclude_columns:
        plot_numeric_statistics(resistant_pancreatic_df, column, 'Resistant_Pancreatic_Cancer')

```


```{r determine cutpoints}

# Define cutoff points for Albumin, LMR, PLT, LY, ANC, NLR, Alk_Phos, and Prognostic_Score_ALAN
cutoff_points <- list(
  Albumin = 3.5,
  LMR = 2.1,
  PLT = 300,
  LY = 1.5,
  MON = 0.8,
  ANC = c(4, 8),
  NLR = c(3, 5),
  Alk_Phos = c(135, 200),
  Prognostic_Score_ALAN = c(0, 2, 4),
  Age = c(60, 65, 70)
)

# Function to categorize values based on cutoff points
categorize_values <- function(df) {
  for (variable in names(cutoff_points)) {
    if (variable %in% colnames(df)) {
      if (variable == "Prognostic_Score_ALAN") {
        df[[paste0(variable, "_category")]] <- cut(df[[variable]], 
                                                    breaks = c(-Inf, 0, 2, Inf),
                                                    labels = c("0", "1-2", "3-4"))
      } else if (is.numeric(cutoff_points[[variable]])) {
        for (cutoff in cutoff_points[[variable]]) {
          category_column <- ifelse(df[[variable]] < cutoff, 
                                    paste0("< ", cutoff), 
                                    paste0(">= ", cutoff))
          df <- cbind(df, category_column)
          colnames(df)[ncol(df)] <- paste0(variable, "_", cutoff)
        }
      } else {
        cutoff <- cutoff_points[[variable]]
        category_column <- cut(df[[variable]], 
                               breaks = c(-Inf, cutoff, Inf),
                               labels = c(paste0("< ", cutoff), 
                                          paste0(">= ", cutoff)))
        df <- cbind(df, category_column)
        colnames(df)[ncol(df)] <- paste0(variable, "_category")
      }
    } else {
      cat(paste("Column '", variable, "' not found in the DataFrame.\n"))
    }
  }
  return(df)
}


# Apply categorization to each DataFrame
categorized_npt_pancreatic_df <- categorize_values(py$npt_pancreatic_df)
categorized_resistant_pancreatic_df <- categorize_values(py$resistant_pancreatic_df)

# Check the result
print("Categorized NPT Pancreatic DataFrame:")
print(head(categorized_npt_pancreatic_df))

print("\nCategorized Resistant Pancreatic DataFrame:")
print(head(categorized_resistant_pancreatic_df))

```
```{r convert to factors}

# Function to convert specified columns to factors
convert_to_factors <- function(df, columns_to_convert) {
    df[, columns_to_convert] <- lapply(df[, columns_to_convert], factor)
    return(df)
}

# Columns to convert to factors
columns_to_convert <- c('Age_60', 'Age_65', 'Age_70', 'Albumin_3.5', 'LMR_2.1', 'PLT_300', 'LY_1.5', 'MON_0.8', 
                        'ANC_4', 'ANC_8', 'NLR_3', 'NLR_5', 'Alk_Phos_135', 'Alk_Phos_200')

# Convert columns to factors for categorized_npt_pancreatic_df
categorized_npt_pancreatic_df <- convert_to_factors(categorized_npt_pancreatic_df, columns_to_convert)

# Convert columns to factors for categorized_resistant_pancreatic_df
categorized_resistant_pancreatic_df <- convert_to_factors(categorized_resistant_pancreatic_df, columns_to_convert)

# Check the structure of the dataframes
str(categorized_npt_pancreatic_df)
str(categorized_resistant_pancreatic_df)

```


## Categoric Summary

Calculate the Categorical statistics for our new pancreatic data frame

```{python calculate categoric}
def calculate_categorical_statistics(data, title="Categorical Statistics"):
    # Check if data is a DataFrame
    if not isinstance(data, pd.DataFrame):
        raise ValueError("Input 'data' must be a pandas DataFrame.")
    
    # Drop the 'ID' column if it exists
    data = data.drop(columns=['ID'], errors='ignore')
    
    # Initialize an empty list to store results
    result_list = []
    
    # Iterate over each non-numeric variable
    for var in data.select_dtypes(exclude=['number']).columns:
        # Get value counts for the current variable
        categories = data[var].value_counts()
        
        # Append the results to the list
        result_list.append(pd.DataFrame({
            'Variable': [var] * len(categories),
            'Levels': categories.index,
            'UniqueValues': len(categories),
            'Frequencies': categories.values.tolist(),
            'Proportions': (categories / categories.sum()).map(lambda x: f"{x:.2%}").tolist()
        }))
    
    # Concatenate the individual DataFrames into one
    result = pd.concat(result_list, ignore_index=True)
    
    # Return result DataFrame
    return result
```

```{python categorized stats}
import warnings

# Suppress FutureWarnings
warnings.filterwarnings("ignore", category=FutureWarning)


# Calling the Python function on the R data frames
categorized_npt_pancreatic_stats = calculate_categorical_statistics(r.categorized_npt_pancreatic_df)
categorized_resistant_pancreatic_stats = calculate_categorical_statistics(r.categorized_resistant_pancreatic_df)

print(categorized_npt_pancreatic_stats)
print(categorized_resistant_pancreatic_stats)


```


```{r advanced categoric summary}

library(gt)

# Define a function to save gt tables as images
save_gt_as_image <- function(table, filename) {
  gtsave(table, filename = filename, path = "output/Pancreatic_Output")
}

# Call the calc_cat_stats function and save the resulting gt tables
cat_stats_npt_pancreatic <- calc_cat_stats(categorized_npt_pancreatic_df, title = "Categoric Statistics for NPT- Pancreatic Cancer")
save_gt_as_image(cat_stats_npt_pancreatic, "categoric_stats_npt_pancreatic.png")

cat_stats_resistant_pancreatic <- calc_cat_stats(categorized_resistant_pancreatic_df, title = "Categoric Statistics for Resistant- Pancreatic Cancer")
save_gt_as_image(cat_stats_resistant_pancreatic, "categoric_stats_resistant_pancreatic.png")


```

```{python categoric distribution}
import seaborn as sns
import matplotlib.pyplot as plt
import numpy as np

def plot_combined_categorical_statistics(data, title="Categorical Statistics"):
    # Create a copy of the data to avoid modifying the original DataFrame
    data_copy = data.copy()
    
    # Remove rows where the split is 100% to 0%
    data_copy = data_copy[(data_copy['Proportions'] != '100.00%') & (data_copy['Proportions'] != '0.00%')]
    
    # Exclude the 'Risk' column
    data_copy = data_copy[data_copy['Variable'] != 'Risk_Group_ALAN']
    
    # Convert Proportions column to numeric
    data_copy['Proportions'] = data_copy['Proportions'].str.rstrip('%').astype(float)
    
    # Combine similar variables
    data_copy['Variable'] = data_copy['Variable'].str.split('_').str[0]  # Extract the part before '_'
    
    # Group by Variable and Levels, calculate mean and standard error of proportions
    grouped_data = data_copy.groupby(['Variable', 'Levels'])['Proportions'].agg(['mean', 'sem']).reset_index()
    
    # Initialize the plot
    sns.set(style="whitegrid")
    plt.figure(figsize=(16, 8))  # Increase figure width
    
    # Create the bar plot
    sns.barplot(data=grouped_data, x='Levels', y='mean', hue='Variable')
    
    # Add error bars
    plt.errorbar(x=np.arange(len(grouped_data['Levels'].unique())), y=grouped_data['mean'], 
                 yerr=grouped_data['sem'], fmt='none', ecolor='black', capsize=3)  # Adjust capsize
    
    # Add labels above each bar
    for index, row in grouped_data.iterrows():
        plt.text(index, row['mean'], f"{row['mean']:.1f}", ha='center', va='bottom', fontsize=6)
    
    # Set title and labels with adjusted font size
    plt.title(title, fontsize=16)
    plt.xlabel('Levels', fontsize=14)
    plt.ylabel('Proportion', fontsize=14)
    plt.xticks(rotation=45, fontsize=8, ha='right')  # Rotate x-axis labels and adjust font size
    plt.yticks(fontsize=8)  # Adjust font size for y-axis labels
    
    # Adjust legend size and position
    plt.legend(title='Variable', fontsize=6, title_fontsize=8, loc='upper right')
    
    # Adjust spacing
    plt.tight_layout()  # Adjust spacing
    
    # Show plot
    plt.show()

# Example usage:
# Plot combined categorical statistics
plot_combined_categorical_statistics(categorized_npt_pancreatic_stats, title="Categorical Statistics for NPT- Pancreatic Cancer")
plot_combined_categorical_statistics(categorized_resistant_pancreatic_stats, title="Categorical Statistics for Resistant- Pancreatic Cancer")

plt.close()  # Close the plot to avoid displaying it again later


```




## Survival Analysis

```{r Suvival Object}

# Create survival object for categorized_resistant_pancreatic_df
surv_obj_resistant <- Surv(time = categorized_resistant_pancreatic_df$time_diff_months, event = categorized_resistant_pancreatic_df$Event_Status)

# Create survival object for categorized_npt_pancreatic_df
surv_obj_npt <- Surv(time = categorized_npt_pancreatic_df$time_diff_months, event = categorized_npt_pancreatic_df$Event_Status)
```

# Overall Kaplan Meier 

```{r Kaplan Meier}

# Load required libraries
library(survival)
library(survminer)
library(ggsci)
library(ggsurvfit)
library(ggplotify)

# Check if 'ggsurvplot' is loaded in the namespace
  if (!"ggsurvplot" %in% loadedNamespaces()) {
    library(survminer)
  }
resistant_fit <- survfit(Surv(time = categorized_resistant_pancreatic_df$time_diff_months, event = categorized_resistant_pancreatic_df$Event_Status) ~1, data = categorized_resistant_pancreatic_df)

resistant_kmplot <- ggsurvplot(resistant_fit,
                              data = categorized_resistant_pancreatic_df,
                              title = "Survival Curve for Resistant Pancreatic Cancer",
                              censor = TRUE,
                              xlab = "Time (Months)",
                              ylab = "Survival Probability",
                              conf.int = TRUE,
                              conf.int.style = "step",
                              conf.int.alpha = 0.2,
                              ggtheme = theme_minimal(),
                              surv.median.line = "hv",
                              xlim = c(0, 24),
                              break.time.by = 3,
                              breaks = seq(0, 24, by = 3),
                              surv.scale = "percent",
                              legend.labs = paste("Resistant- Pancreatic Cancer (N =", nrow(categorized_resistant_pancreatic_df),")"),
                              palette = "lancet")

resistant_kmplot <- resistant_kmplot + ggsurvfit::theme_ggsurvfit_KMunicate()



npt_fit <- survfit(Surv(time = categorized_npt_pancreatic_df$time_diff_months, event = categorized_npt_pancreatic_df$Event_Status) ~1, data = categorized_npt_pancreatic_df)

npt_kmplot <- ggsurvplot(npt_fit,
                              data = categorized_npt_pancreatic_df,
                              title = "Survival Curve for No Prior Treatment Pancreatic Cancer",
                              censor = TRUE,
                              xlab = "Time (Months)",
                              ylab = "Survival Probability",
                              conf.int = TRUE,
                              conf.int.style = "step",
                              conf.int.alpha = 0.2,
                              ggtheme = theme_minimal(),
                              surv.median.line = "hv",
                              xlim = c(0, 24),
                              break.time.by = 3,
                              breaks = seq(0, 24, by = 3),
                              surv.scale = "percent",
                              legend.labs = paste("NPT- Pancreatic Cancer (N =", nrow(categorized_npt_pancreatic_df),")"),
                              palette = "lancet")

npt_kmplot <- npt_kmplot + ggsurvfit::theme_ggsurvfit_KMunicate()

resistant_kmplot
npt_kmplot


# Convert resistant_kmplot and npt_kmplot to ggplot objects
resistant_kmplot_gg <- resistant_kmplot$plot
npt_kmplot_gg <- npt_kmplot$plot

# Save the ggplot objects
ggsave(filename = "output/Pancreatic_Output/resistant_kmplot.png", plot = resistant_kmplot_gg, width = 10, height = 6)
ggsave(filename = "output/Pancreatic_Output/npt_kmplot.png", plot = npt_kmplot_gg, width = 10, height = 6)


```

```{r combined KM overall}
combined_fit <- list(NPT = npt_fit, Resistant = resistant_fit)
combined_kmplot <- ggsurvplot_combine(combined_fit,
                                  data = data_resistant,
                                  title = "Survival Curve for Resistant- vs. NPT- Pancreatic Cancer",
                                  censor = TRUE,
                                  xlab = "Time (Months)",
                                  ylab = "Survival Probability",
                                  conf.int = TRUE,
                                  conf.int.style = "step",
                                  conf.int.alpha = 0.2,
                                  ggtheme = theme_minimal(),
                                  surv.median.line = "hv",
                                  xlim = c(0, 24),
                                  break.time.by = 3,
                                  breaks = seq(0, 24, by = 3),
                                  palette = "lancet")
combined_kmplot <- combined_kmplot + ggsurvfit::theme_ggsurvfit_KMunicate()

combined_kmplot
  
combined_gg <- combined_kmplot$plot
# Save the ggplot objects
ggsave(filename = "output/Pancreatic_Output/resistant_v_npt_kmplot.png", plot = combined_gg, width = 10, height = 6)
```


```{r rename df}
npt_df <- categorized_npt_pancreatic_df
resistant_df <- categorized_resistant_pancreatic_df

colnames(npt_df)
```

## Km Fit Curve



```{r Resistant KMFit}
# Define column names, variables, and cutoffs
column_names <- c("Albumin_3.5", "LMR_2.1", "MON_0.8", "LY_1.5", "ANC_4", "ANC_8", "NLR_3", "NLR_5", "PLT_300", "Alk_Phos_135", "Alk_Phos_200", "Age_60", "Age_65", "Age_70", "Prognostic_Score_ALAN_category")

# Initialize a list to store survival fits for resistant Pancreatic Cancer 
resistant_km_fits <- list()

# Loop through variables for resistant Pancreatic Cancer
for (col in column_names) {
  # Construct formula with variable name extracted from column name
  formula <- as.formula(paste("surv_obj_resistant ~", col))
  
  # Fit Kaplan-Meier survival curve
  resistant_km_fit <- survfit(formula, data = resistant_df)
  
  # Store the fit in the list with a descriptive name
  resistant_km_fits[[paste("resistant_km_fit_", col, sep = "")]] <- resistant_km_fit
}

# Access results using names like resistant_km_fit_Albumin_3.5 etc.
print("Resistant Pancreatic Cancer Survival Fits:")
print(resistant_km_fits)

# Add a line of dashes for separation
cat("\n", paste(rep("-", 40), collapse = ""), "\n")

```
```{r NPT KM Fit}
# Define column names, variables, and cutoffs
column_names <- c("Albumin_3.5", "LMR_2.1", "MON_0.8", "LY_1.5", "ANC_4", "ANC_8", "NLR_3", "NLR_5", "PLT_300", "Alk_Phos_135", "Alk_Phos_200", "Age_60", "Age_65", "Age_70", "Prognostic_Score_ALAN_category")

# Initialize a list to store survival fits for NPT Pancreatic Cancer 
npt_km_fits <- list()

# Loop through variables for NPT Pancreatic Cancer
for (col in column_names) {
  # Construct formula with variable name extracted from column name
  formula <- as.formula(paste("surv_obj_npt ~", col))
  
  # Fit Kaplan-Meier survival curve
  npt_km_fit <- survfit(formula, data = npt_df)
  
  # Store the fit in the list with a descriptive name
  npt_km_fits[[paste("npt_km_fit_", col, sep = "")]] <- npt_km_fit
}

# Access results using names like resistant_km_fit_Albumin_3.5 etc.
print("NPT- Pancreatic Cancer Survival Fits:")
print(npt_km_fits)

# Add a line of dashes for separation
cat("\n", paste(rep("-", 40), collapse = ""), "\n")


```
##LogRank

For Resistant Log-Rank

1. **Albumin_3.5**:
   - There's a significant difference in survival between patients with Albumin levels below 3.5 and those with levels equal to or above 3.5 (p = 2e-04). Patients with Albumin levels below 3.5 have a higher observed-to-expected ratio than those with levels equal to or above 3.5.

2. **LMR_2.1**:
   - There's no significant difference in survival between patients with LMR levels below 2.1 and those with levels equal to or above 2.1 (p = 0.09).

3. **MON_0.8**:
   - There's no significant difference in survival between patients with MON levels below 0.8 and those with levels equal to or above 0.8 (p = 0.08).

4. **LY_1.5**:
   - There's no significant difference in survival between patients with LY levels below 1.5 and those with levels equal to or above 1.5 (p = 0.2).

5. **ANC_4**:
   - There's no significant difference in survival between patients with ANC levels below 4 and those with levels equal to or above 4 (p = 0.1).

6. **ANC_8**:
   - There's a marginally significant difference in survival between patients with ANC levels below 8 and those with levels equal to or above 8 (p = 0.07). Patients with ANC levels below 8 have a higher observed-to-expected ratio than those with levels equal to or above 8.

7. **NLR_3**:
   - There's a significant difference in survival between patients with NLR levels below 3 and those with levels equal to or above 3 (p = 0.02). Patients with NLR levels below 3 have a higher observed-to-expected ratio than those with levels equal to or above 3.

8. **NLR_5**:
   - There's a significant difference in survival between patients with NLR levels below 5 and those with levels equal to or above 5 (p = 0.02). Patients with NLR levels below 5 have a higher observed-to-expected ratio than those with levels equal to or above 5.

9. **PLT_300**:
   - There's no significant difference in survival between patients with PLT levels below 300 and those with levels equal to or above 300 (p = 0.1).

10. **Alk_Phos_135**:
    - There's a significant difference in survival between patients with Alk_Phos levels below 135 and those with levels equal to or above 135 (p = 0.008). Patients with Alk_Phos levels below 135 have a higher observed-to-expected ratio than those with levels equal to or above 135.

11. **Alk_Phos_200**:
    - There's a marginally significant difference in survival between patients with Alk_Phos levels below 200 and those with levels equal to or above 200 (p = 0.06).

12. **Age_60, Age_65, Age_70**:
    - There's no significant difference in survival between patients in different age groups (p > 0.05 for all comparisons).

13. **Prognostic_Score_ALAN_category**:
    - There's a significant difference in survival among patients in different prognostic score categories (p = 0.001). Post-hoc pairwise comparisons show significant differences between all categories.

These results suggest that factors like Albumin, ANC, NLR, and Alk_Phos levels might be associated with survival outcomes in patients with resistant pancreatic cancer, while other factors like LMR, MON, LY, PLT, and age may not have significant associations.

```{r Resistant Log-Rank Test}
 
# Define column names
column_names <- c("Albumin_3.5", "LMR_2.1", "MON_0.8", "LY_1.5", "ANC_4", "ANC_8", "NLR_3", "NLR_5", "PLT_300", "Alk_Phos_135", "Alk_Phos_200", "Age_60", "Age_65", "Age_70", "Prognostic_Score_ALAN_category")

# Initialize an empty data frame to store log-rank test results for Resistant Pancreatic Cancer
log_rank_results_df_resistant <- data.frame(
  variable = character(),
  cutoff = numeric(),
  logrank_statistic = numeric(),
  logrank_p_value = numeric(),
  stringsAsFactors = FALSE
)

# Loop through variables for resistant Pancreatic Cancer
for (col in column_names) {
  if (grepl("^Prognostic_Score_ALAN_Category", col)) {
    # Treat categorical variable as a factor
    formula <- as.formula(paste("surv_obj_resistant ~ factor(", col, ")"))
    
    # Perform log-rank test
    resistant_logrank <- survdiff(formula, data = resistant_df)
    
    # Store log-rank test results in data frame
    log_rank_results_df_resistant <- rbind(log_rank_results_df_resistant, data.frame(
      variable = col,
      cutoff = "N/A",
      logrank_statistic = resistant_logrank$chisq,
      logrank_p_value = 1 - pchisq(resistant_logrank$chisq, df = 1),
      stringsAsFactors = FALSE
    ))
    
    # Print log-rank test information for Resistant Pancreatic Cancer
    cat(rep("-", 20), "\n")
    cat("Log-rank tests for Resistant Pancreatic Cancer -", col, "\n")
    cat(rep("-", 20), "\n")
    print(resistant_logrank)
  } else {
    # Extract cutoff from column name using regular expression
    cutoff <- as.numeric(sub("^.*_(\\d+(\\.\\d+)?)$", "\\1", col))
    formula <- as.formula(paste("surv_obj_resistant ~", col))
  
    # Perform log-rank test
    resistant_logrank <- survdiff(formula, data = resistant_df)
  
    # Store log-rank test results in data frame
    log_rank_results_df_resistant <- rbind(log_rank_results_df_resistant, data.frame(
      variable = col,
      cutoff = cutoff,
      logrank_statistic = resistant_logrank$chisq,
      logrank_p_value = 1 - pchisq(resistant_logrank$chisq, df = 1),
      stringsAsFactors = FALSE
    ))
  
    # Print log-rank test information for Resistant Pancreatic Cancer
    cat(rep("-", 20), "\n")
    cat("Log-rank tests for Resistant Pancreatic Cancer -", col, "\n")
    cat(rep("-", 20), "\n")
    print(resistant_logrank)
  }
}


# Display log-rank test results in a table
kable(log_rank_results_df_resistant, caption = "Log-rank Test Results for Resistant Pancreatic Cancer")

```
For NPT Log-Rank: 

1. **Albumin_3.5**:
   - There's a significant difference in survival between patients with Albumin levels below 3.5 and those with levels equal to or above 3.5 (p = 6e-04). Patients with Albumin levels below 3.5 have a higher observed-to-expected ratio than those with levels equal to or above 3.5.

2. **LMR_2.1**:
   - There's no significant difference in survival between patients with LMR levels below 2.1 and those with levels equal to or above 2.1 (p = 0.07).

3. **MON_0.8**:
   - There's a significant difference in survival between patients with MON levels below 0.8 and those with levels equal to or above 0.8 (p = 0.005). Patients with MON levels below 0.8 have a higher observed-to-expected ratio than those with levels equal to or above 0.8.

4. **LY_1.5**:
   - There's no significant difference in survival between patients with LY levels below 1.5 and those with levels equal to or above 1.5 (p = 0.4).

5. **ANC_4**:
   - There's no significant difference in survival between patients with ANC levels below 4 and those with levels equal to or above 4 (p = 0.2).

6. **ANC_8**:
   - There's a significant difference in survival between patients with ANC levels below 8 and those with levels equal to or above 8 (p = 0.03). Patients with ANC levels below 8 have a higher observed-to-expected ratio than those with levels equal to or above 8.

7. **NLR_3**:
   - There's no significant difference in survival between patients with NLR levels below 3 and those with levels equal to or above 3 (p = 0.3).

8. **NLR_5**:
   - There's no significant difference in survival between patients with NLR levels below 5 and those with levels equal to or above 5 (p = 0.2).

9. **PLT_300**:
   - There's no significant difference in survival between patients with PLT levels below 300 and those with levels equal to or above 300 (p = 0.9).

10. **Alk_Phos_135**:
   - There's no significant difference in survival between patients with Alk_Phos levels below 135 and those with levels equal to or above 135 (p = 0.4).

11. **Alk_Phos_200**:
   - There's no significant difference in survival between patients with Alk_Phos levels below 200 and those with levels equal to or above 200 (p = 0.4).

12. **Age_60, Age_65, Age_70**:
   - There's no significant difference in survival between patients in different age groups (p > 0.05 for all comparisons).

13. **Prognostic_Score_ALAN_category**:
   - There's a significant difference in survival among patients in different prognostic score categories (p = 8e-09). Post-hoc pairwise comparisons show significant differences between all categories.

Overall, these results provide insights into the potential prognostic factors for survival in the "npt" dataset. Factors like Albumin, MON, and ANC levels seem to have significant associations with survival outcomes, while others like LMR, LY, NLR, PLT, Alk_Phos, and age may not be strongly associated.


```{r NPT Log Rank}
# Define column names for npt
column_names_npt <- c("Albumin_3.5", "LMR_2.1", "MON_0.8", "LY_1.5", "ANC_4", "ANC_8", "NLR_3", "NLR_5", "PLT_300", "Alk_Phos_135", "Alk_Phos_200", "Age_60", "Age_65", "Age_70", "Prognostic_Score_ALAN_category")

# Initialize an empty data frame to store log-rank test results for npt
log_rank_results_df_npt <- data.frame(
  variable = character(),
  cutoff = numeric(),
  logrank_statistic = numeric(),
  logrank_p_value = numeric(),
  stringsAsFactors = FALSE
)

# Loop through variables for npt
for (col in column_names_npt) {
  if (grepl("^Prognostic_Score_ALAN_Category", col)) {
    # Treat categorical variable as a factor
    formula <- as.formula(paste("surv_obj_npt ~ factor(", col, ")"))
    
    # Perform log-rank test
    npt_logrank <- survdiff(formula, data = npt_df)
    
    # Store log-rank test results in data frame
    log_rank_results_df_npt <- rbind(log_rank_results_df_npt, data.frame(
      variable = col,
      cutoff = "N/A",
      logrank_statistic = npt_logrank$chisq,
      logrank_p_value = 1 - pchisq(npt_logrank$chisq, df = 1),
      stringsAsFactors = FALSE
    ))
    
    # Print log-rank test information for npt
    cat(rep("-", 20), "\n")
    cat("Log-rank tests for npt -", col, "\n")
    cat(rep("-", 20), "\n")
    print(npt_logrank)
  } else {
    # Extract cutoff from column name using regular expression
    cutoff <- as.numeric(sub("^.*_(\\d+(\\.\\d+)?)$", "\\1", col))
    formula <- as.formula(paste("surv_obj_npt ~", col))
  
    # Perform log-rank test
    npt_logrank <- survdiff(formula, data = npt_df)
  
    # Store log-rank test results in data frame
    log_rank_results_df_npt <- rbind(log_rank_results_df_npt, data.frame(
      variable = col,
      cutoff = cutoff,
      logrank_statistic = npt_logrank$chisq,
      logrank_p_value = 1 - pchisq(npt_logrank$chisq, df = 1),
      stringsAsFactors = FALSE
    ))
  
    # Print log-rank test information for npt
    cat(rep("-", 20), "\n")
    cat("Log-rank tests for npt -", col, "\n")
    cat(rep("-", 20), "\n")
    print(npt_logrank)
  }
}

# Display log-rank test results for npt in a table
kable(log_rank_results_df_npt, caption = "Log-rank Test Results for npt- pancreatic cancer")


```
### Pairwise LogRank

These results are pairwise log-rank tests comparing different levels of the variable "Prognostic_Score_ALAN_category" within the "Resistant" data. Let's interpret each pairwise comparison:

1. **Pairwise log-rank test between 1 and 2:**
   - Chisq: 6.2 on 1 degree of freedom, p-value = 0.01
   - Interpretation: There is a statistically significant difference in survival between patients with a prognostic score of 1 and those with a score of 2.

2. **Pairwise log-rank test between 1 and 3:**
   - Chisq: 9.8 on 1 degree of freedom, p-value = 0.002
   - Interpretation: There is a statistically significant difference in survival between patients with a prognostic score of 1 and those with a score of 3-4.

3. **Pairwise log-rank test between 2 and 3:**
   - Chisq: 3.8 on 1 degree of freedom, p-value = 0.05
   - Interpretation: There is a marginally significant difference in survival between patients with a prognostic score of 2 and those with a score of 3-4.

In summary, these results indicate that the prognostic score categories have significant or marginally significant differences in survival among the patients with resistant pancreatic cancer.

```{r Resistant Pairwise Logrank}
# Get unique levels of Prognostic_Score_ALAN_category
levels_resistant <- unique(resistant_df$Prognostic_Score_ALAN_category)

# Initialize a list to store pairwise log-rank test results
pairwise_results_resistant <- list()

# Perform pairwise log-rank tests
for (i in 1:(length(levels_resistant)-1)) {
  for (j in (i+1):length(levels_resistant)) {
    level1 <- levels_resistant[i]
    level2 <- levels_resistant[j]
    cat("Pairwise log-rank test between", level1, "and", level2, "\n")
    formula <- as.formula(paste("Surv(time_diff_months, Event_Status) ~ Prognostic_Score_ALAN_category"))
    pairwise_test <- survdiff(formula, subset(resistant_df, Prognostic_Score_ALAN_category %in% c(level1, level2)))
    print(pairwise_test)
    cat("\n")
    # Store the pairwise test result
    pairwise_results_resistant[[paste("pairwise_test_", level1, "_vs_", level2, sep = "")]] <- pairwise_test
  }
}

# Print the results
print("Pairwise log-rank test results for resistant pancreatic cancer:")
print(pairwise_results_resistant)

```


For NPT Pairwise

The pairwise log-rank test results provide information about the differences in survival distributions between different categories of the `Prognostic_Score_ALAN_category` variable in the "npt" dataset. Here's what the results tell us:

1. Pairwise test between levels 2 and 1:
   - The chi-square statistic is 0, indicating that there is no significant difference in survival distributions between category 2 and category 1 of the `Prognostic_Score_ALAN_category` variable.
   - The p-value is 0.9, suggesting that there is no evidence to reject the null hypothesis of no difference in survival distributions between the two categories.

2. Pairwise test between levels 2 and 3:
   - The chi-square statistic is 22.4, indicating a significant difference in survival distributions between category 2 and category 3 of the `Prognostic_Score_ALAN_category` variable.
   - The p-value is very small (2e-06), suggesting strong evidence to reject the null hypothesis of no difference in survival distributions between the two categories.

3. Pairwise test between levels 1 and 3:
   - The chi-square statistic is 21.9, indicating a significant difference in survival distributions between category 1 and category 3 of the `Prognostic_Score_ALAN_category` variable.
   - The p-value is very small (3e-06), suggesting strong evidence to reject the null hypothesis of no difference in survival distributions between the two categories.

In summary, these pairwise comparisons reveal significant differences in survival distributions between certain categories of the `Prognostic_Score_ALAN_category` variable in the "npt" dataset. This information could be valuable for understanding the impact of different prognostic scores on survival outcomes in the dataset.



```{r NPT Pairwise Logrank}

# Get unique levels of Prognostic_Score_ALAN_category
levels_npt <- unique(npt_df$Prognostic_Score_ALAN_category)

# Initialize a list to store pairwise log-rank test results
pairwise_results_npt <- list()

# Perform pairwise log-rank tests
for (i in 1:(length(levels_npt)-1)) {
  for (j in (i+1):length(levels_npt)) {
    level1 <- levels_npt[i]
    level2 <- levels_npt[j]
    cat("Pairwise log-rank test between", level1, "and", level2, "\n")
    formula <- as.formula(paste("Surv(time_diff_months, Event_Status) ~ Prognostic_Score_ALAN_category"))
    pairwise_test <- survdiff(formula, subset(npt_df, Prognostic_Score_ALAN_category %in% c(level1, level2)))
    print(pairwise_test)
    cat("\n")
    # Store the pairwise test result
    pairwise_results_npt[[paste("pairwise_test_", level1, "_vs_", level2, sep = "")]] <- pairwise_test
  }
}

# Print the results
print("Pairwise log-rank test results for npt- pancreatic cancer:")
print(pairwise_results_npt)


```

##Cox Proportional Hazards

```{r Resistant Cox Proportional Hazards}
# Initialize lists to store Cox models, p-values, and hazard ratios for Resistant Pancreatic Cancer
cox_p_values_list_resistant <- list()
cox_hazard_ratios_list_resistant <- list()

# Loop through variables for Resistant Pancreatic Cancer
for (col in column_names) {
  
   # Extract cutoff and variable name using regular expressions
  cutoff <- as.numeric(sub("^.*_(\\d+(\\.\\d+)?)$", "\\1", col))
  variable <- sub("^(.*)_\\d+(\\.\\d+)?$", "\\1", col)
  
  # Create formula for Cox model
  formula_resistant <- as.formula(paste("Surv(time_diff_months, Event_Status) ~", col))
    
    # Fit Cox model
    cox_model_resistant <- coxph(formula_resistant, data = resistant_df)
    
    # Print Cox model for Resistant Pancreatic Cancer
    cat(rep("-", 30), "\n")
    cat("Cox Proportional Hazards for Resistant Pancreatic Cancer  -", col, "\n")
    cat(rep("-", 30), "\n")
    print(cox_model_resistant)
    
    # Create properly formatted column name for p-value extraction
    coef_name_resistant <- paste(col, ">= ", cutoff, sep = "")
    
    # Check if the variable is Prognostic_Score_ALAN_category
    if (variable == "Prognostic_Score_ALAN_category") {
      # Treat categorical variable as a factor
      formula_resistant <- as.formula(paste("Surv(time_diff_months, Event_Status) ~ factor(", col, ")"))
      
      # Perform Cox model for Prognostic_Score_ALAN_category
      cox_model_resistant <- coxph(formula_resistant, data = resistant_df)
      
      # Extract p-value
      cox_p_value_resistant <- as.numeric(format(summary(cox_model_resistant)$coefficients[, "Pr(>|z|)"], scientific = TRUE, digits = 3))
      
      # Print p-value for Resistant Pancreatic Cancer
      cat("P-value:", cox_p_value_resistant, "\n")
    } else {
      # Extract p-value
      cox_p_value_resistant <- as.numeric(format(summary(cox_model_resistant)$coefficients[coef_name_resistant, "Pr(>|z|)"], scientific = TRUE, digits = 3))
      
      # Print p-value for Resistant Pancreatic Cancer
      cat("P-value:", cox_p_value_resistant, "\n")
    }
    
    # Extract Hazard Ratio
    cox_hazard_ratio_resistant <- exp(coef(cox_model_resistant))
    
    # Print Hazard Ratio for Resistant Pancreatic Cancer
    cat("Hazard Ratio:", cox_hazard_ratio_resistant, "\n")
    
    # Append the results to respective lists for Resistant Pancreatic Cancer
    cox_p_values_list_resistant[[length(cox_p_values_list_resistant) + 1]] <- c(col, cutoff, cox_p_value_resistant)
    cox_hazard_ratios_list_resistant[[length(cox_hazard_ratios_list_resistant) + 1]] <- c(col, cutoff, cox_hazard_ratio_resistant)
    
}

# Convert lists to data frames for Resistant Pancreatic Cancer
cox_p_values_df_resistant <- as.data.frame(do.call(rbind, cox_p_values_list_resistant), stringsAsFactors = FALSE)
colnames(cox_p_values_df_resistant) <- c("column_names", "cutoff", "cox_p_value")
cox_p_values_df_resistant$cutoff <- as.numeric(cox_p_values_df_resistant$cutoff)
cox_p_values_df_resistant$cox_p_value <- as.numeric(cox_p_values_df_resistant$cox_p_value)

cox_hazard_ratios_df_resistant <- as.data.frame(do.call(rbind, cox_hazard_ratios_list_resistant), stringsAsFactors = FALSE)
colnames(cox_hazard_ratios_df_resistant) <- c("column_names", "cutoff", "cox_hazard_ratio")
cox_hazard_ratios_df_resistant$cutoff <- as.numeric(cox_hazard_ratios_df_resistant$cutoff)
cox_hazard_ratios_df_resistant$cox_hazard_ratio <- as.numeric(cox_hazard_ratios_df_resistant$cox_hazard_ratio)

# Merge the two data frames for Resistant Pancreatic Cancer
coxph_df_resistant <- merge(cox_p_values_df_resistant, cox_hazard_ratios_df_resistant, by = c("column_names", "cutoff"), sort = FALSE)

# Select only the relevant columns
coxph_df_resistant <- coxph_df_resistant[, c("column_names", "cutoff", "cox_p_value", "cox_hazard_ratio")]

# Print the combined data frame for Resistant Pancreatic Cancer using kable
kable(coxph_df_resistant, caption = "Cox Proportional Hazards Results for Resistant Pancreatic Cancer")

# Print the structure of the combined data frame for Resistant Pancreatic Cancer
str(coxph_df_resistant)

```

```{r NPT Cox Proportional Hazard}
# Initialize lists to store Cox models, p-values, and hazard ratios for NPT
cox_p_values_list_npt <- list()
cox_hazard_ratios_list_npt <- list()

# Loop through variables for NPT
for (col in column_names) {
  
   # Extract cutoff and variable name using regular expressions
  cutoff <- as.numeric(sub("^.*_(\\d+(\\.\\d+)?)$", "\\1", col))
  variable <- sub("^(.*)_\\d+(\\.\\d+)?$", "\\1", col)
  
  # Create formula for Cox model
  formula_npt <- as.formula(paste("Surv(time_diff_months, Event_Status) ~", col))
    
    # Fit Cox model
    cox_model_npt <- coxph(formula_npt, data = npt_df)
    
    # Print Cox model for NPT
    cat(rep("-", 30), "\n")
    cat("Cox Proportional Hazards for NPT -", col, "\n")
    cat(rep("-", 30), "\n")
    print(cox_model_npt)
    
    # Create properly formatted column name for p-value extraction
    coef_name_npt <- paste(col, ">= ", cutoff, sep = "")
    
    # Check if the variable is Prognostic_Score_ALAN_category
    if (variable == "Prognostic_Score_ALAN_category") {
      # Treat categorical variable as a factor
      formula_npt <- as.formula(paste("Surv(time_diff_months, Event_Status) ~ factor(", col, ")"))
      
      # Perform Cox model for Prognostic_Score_ALAN_category
      cox_model_npt <- coxph(formula_npt, data = npt_df)
      
      # Extract p-value
      cox_p_value_npt <- as.numeric(format(summary(cox_model_npt)$coefficients[, "Pr(>|z|)"], scientific = TRUE, digits = 3))
      
      # Print p-value for NPT
      cat("P-value:", cox_p_value_npt, "\n")
    } else {
      # Extract p-value
      cox_p_value_npt <- as.numeric(format(summary(cox_model_npt)$coefficients[coef_name_npt, "Pr(>|z|)"], scientific = TRUE, digits = 3))
      
      # Print p-value for NPT
      cat("P-value:", cox_p_value_npt, "\n")
    }
    
    # Extract Hazard Ratio
    cox_hazard_ratio_npt <- exp(coef(cox_model_npt))
    
    # Print Hazard Ratio for NPT
    cat("Hazard Ratio:", cox_hazard_ratio_npt, "\n")
    
    # Append the results to respective lists for NPT
    cox_p_values_list_npt[[length(cox_p_values_list_npt) + 1]] <- c(col, cutoff, cox_p_value_npt)
    cox_hazard_ratios_list_npt[[length(cox_hazard_ratios_list_npt) + 1]] <- c(col, cutoff, cox_hazard_ratio_npt)
    
}

# Convert lists to data frames for NPT
cox_p_values_df_npt <- as.data.frame(do.call(rbind, cox_p_values_list_npt), stringsAsFactors = FALSE)
colnames(cox_p_values_df_npt) <- c("column_names", "cutoff", "cox_p_value")
cox_p_values_df_npt$cutoff <- as.numeric(cox_p_values_df_npt$cutoff)
cox_p_values_df_npt$cox_p_value <- as.numeric(cox_p_values_df_npt$cox_p_value)

cox_hazard_ratios_df_npt <- as.data.frame(do.call(rbind, cox_hazard_ratios_list_npt), stringsAsFactors = FALSE)
colnames(cox_hazard_ratios_df_npt) <- c("column_names", "cutoff", "cox_hazard_ratio")
cox_hazard_ratios_df_npt$cutoff <- as.numeric(cox_hazard_ratios_df_npt$cutoff)
cox_hazard_ratios_df_npt$cox_hazard_ratio <- as.numeric(cox_hazard_ratios_df_npt$cox_hazard_ratio)

# Merge the two data frames for NPT
coxph_df_npt <- merge(cox_p_values_df_npt, cox_hazard_ratios_df_npt, by = c("column_names", "cutoff"), sort = FALSE)

# Select only the relevant columns
coxph_df_npt <- coxph_df_npt[, c("column_names", "cutoff", "cox_p_value", "cox_hazard_ratio")]

# Print the combined data frame for NPT using kable
kable(coxph_df_npt, caption = "Cox Proportional Hazards Results for NPT- Pancreatic Cancer")

# Print the structure of the combined data frame for NPT
str(coxph_df_npt)


```




##Schoenfeld Residuals Test

```{r Resistant Schoenfeld Test}
# Initialize lists to store Schoenfeld test results and plots for Resistant Pancreatic Cancer
schoenfeld_results_list_resistant <- list()
schoenfeld_plots_list_resistant <- list()

# Loop through variables for Resistant Pancreatic Cancer
for (col in column_names) {
  # Create formula for Cox model
  formula_resistant <- as.formula(paste("Surv(time_diff_months, Event_Status) ~", col))
  
  # Fit Cox model for Resistant Pancreatic Cancer
  cox_model_resistant <- coxph(formula_resistant, data = resistant_df)
  
  # Perform Schoenfeld test for Resistant Pancreatic Cancer
  schoenfeld_test_resistant <- cox.zph(cox_model_resistant)
  
  # Print Schoenfeld test results for Resistant Pancreatic Cancer
  cat(rep("-", 45), "\n")
  cat("Schoenfeld Test for Resistant Pancreatic Cancer -", col, "\n")
  cat(rep("-", 45), "\n")
  print(schoenfeld_test_resistant)
  
  # Store Schoenfeld test result for Resistant Pancreatic Cancer in the list
  schoenfeld_results_list_resistant[[paste("schoenfeld_test", tolower(col), sep = "_")]] <- schoenfeld_test_resistant
  
  # Plot Schoenfeld residuals using ggcoxzph for Resistant Pancreatic Cancer
  schoenfeld_plot_resistant <- ggcoxzph(schoenfeld_test_resistant, caption = paste("Schoenfeld Plot of Resistant Pancreatic Cancer for residuals of", col))
  
  # Store Schoenfeld plot for Resistant Pancreatic Cancer in the list
  schoenfeld_plots_list_resistant[[paste("schoenfeld_plot", tolower(col), sep = "_")]] <- schoenfeld_plot_resistant
  
  # Print the plot for Resistant Pancreatic Cancer
  print(schoenfeld_plot_resistant)
}

# Access results using names like schoenfeld_test_ly etc. for Resistant Pancreatic Cancer
print(schoenfeld_results_list_resistant)
print(schoenfeld_plots_list_resistant)

```


```{r NPT Schoenfeld Test}

# Initialize lists to store Schoenfeld test results and plots for No Prior Treatment Pancreatic Cancer (NPT)
schoenfeld_results_list_npt <- list()
schoenfeld_plots_list_npt <- list()

# Loop through variables for NPT
for (col in column_names) {
  # Create formula for Cox model
  formula_npt <- as.formula(paste("Surv(time_diff_months, Event_Status) ~", col))
  
  # Fit Cox model for NPT
  cox_model_npt <- coxph(formula_npt, data = npt_df)
  
  # Perform Schoenfeld test for NPT
  schoenfeld_test_npt <- cox.zph(cox_model_npt)
  
  # Print Schoenfeld test results for NPT
  cat(rep("-", 45), "\n")
  cat("Schoenfeld Test for NPT- Pancreatic Cancer", col, "\n")
  cat(rep("-", 45), "\n")
  print(schoenfeld_test_npt)
  
  # Store Schoenfeld test result for NPT in the list
  schoenfeld_results_list_npt[[paste("schoenfeld_test", tolower(col), sep = "_")]] <- schoenfeld_test_npt
  
  # Plot Schoenfeld residuals using ggcoxzph for NPT
  schoenfeld_plot_npt <- ggcoxzph(schoenfeld_test_npt, caption = paste("Schoenfeld Plot of No Prior Treatment Pancreatic Cancer for residuals of", col))
  
  # Store Schoenfeld plot for NPT in the list
  schoenfeld_plots_list_npt[[paste("schoenfeld_plot", tolower(col), sep = "_")]] <- schoenfeld_plot_npt
  
  # Print the plot for NPT
  print(schoenfeld_plot_npt)
}

# Access results using names like schoenfeld_test_ly etc. for NPT
print(schoenfeld_results_list_npt)
print(schoenfeld_plots_list_npt)


```


#KM Plots

```{r legend labels}
# Define the column names
column_names <- c("Albumin_3.5", "LMR_2.1", "MON_0.8", "LY_1.5", "ANC_4", "ANC_8", "NLR_3", "NLR_5", "PLT_300", "Alk_Phos_135", "Alk_Phos_200", "Age_60", "Age_65", "Age_70", "Prognostic_Score_ALAN_category")

# Create a separate list of titles for legend labels
legend_titles <- c("Albumin ", "LMR ", "MON ", "LY ", "ANC ", "ANC ", "NLR ", "NLR ", "PLT ", "Alk_Phos ", "Alk_Phos ", "Age ", "Age ", "Age ", "ALAN_Score:")



# Function to create legend labels for each variable
create_legend_labels <- function(df, column_names, legend_titles) {
  legend_labels <- list()
  for (i in seq_along(column_names)) {
    variable <- column_names[i]
    title <- legend_titles[i]
    levels <- levels(df[[variable]])
    if (length(levels) == 2) {
      label1 <- paste(title, levels[1], " (N =", sum(df[[variable]] == levels[1]), ")", sep = "")
      label2 <- paste(title, levels[2], " (N =", sum(df[[variable]] == levels[2]), ")", sep = "")
      legend_labels[[variable]] <- c(label1, label2)
    } else if (length(levels) == 3) {
      label1 <- paste(title, levels[1], " (N =", sum(df[[variable]] == levels[1]), ")", sep = "")
      label2 <- paste(title, levels[2], " (N =", sum(df[[variable]] == levels[2]), ")", sep = "")
      label3 <- paste(title, levels[3], " (N =", sum(df[[variable]] == levels[3]), ")", sep = "")
      legend_labels[[variable]] <- c(label1, label2, label3)
    } else {
      labels <- paste(title, levels, " (N =", table(df[[variable]]), ")", sep = "")
      legend_labels[[variable]] <- labels
    }
  }
  return(legend_labels)
}


# Call the function to create legend labels for each variable
legend_labels_resistant <- create_legend_labels(resistant_df, column_names, legend_titles)

legend_labels_npt <- create_legend_labels(npt_df, column_names, legend_titles) 


print(legend_labels_resistant)
print(legend_labels_npt)
```
```{r create our models}
# Define a function to fit Cox models for each variable in a data frame
fit_cox_models <- function(data, column_names) {
  cox_models <- list()
  for (col in column_names) {
    formula <- as.formula(paste("Surv(time_diff_months, Event_Status) ~", col))
    cox_model <- coxph(formula, data = data)
    cox_models[[col]] <- cox_model
  }
  return(cox_models)
}

# Define a function to fit Kaplan-Meier models for each variable in a data frame
fit_km_models <- function(data, column_names) {
  km_models <- list()
  for (col in column_names) {
    formula <- as.formula(paste("Surv(time_diff_months, Event_Status) ~", col))
    km_model <- survfit(formula, data = data)
    km_models[[col]] <- km_model
  }
  return(km_models)
}

# Fit Cox models for NPT data frame
cox_models_npt <- fit_cox_models(npt_df, column_names)

# Fit Cox models for Resistant data frame
cox_models_resistant <- fit_cox_models(resistant_df, column_names)

# Fit Kaplan-Meier models for NPT data frame
km_models_npt <- fit_km_models(npt_df, column_names)

# Fit Kaplan-Meier models for Resistant data frame
km_models_resistant <- fit_km_models(resistant_df, column_names)
```

```{r check logrank and HR}
# Print the Log-Rank P-value and Hazard Ratio for each variable
cat(rep("-", 40), "\n")
print("Log-Rank pvalue and HR for NPT Pancreatic Cancer")
cat(rep("-", 40), "\n")
for (col in names(km_models_npt)) {
  logrank_p_npt <- log_rank_results_df_npt$logrank_p_value[log_rank_results_df_npt$variable == col]
  cox_HR_npt <- coxph_df_npt$cox_hazard_ratio[coxph_df_npt$column_names == col]
  cat("Variable:", col, "\n")
  cat("Log-Rank P-value:", logrank_p_npt, "\n")
  cat("Hazard Ratio:", cox_HR_npt, "\n")
}


# Print the Log-Rank P-value and Hazard Ratio for each variable
cat(rep("-", 40), "\n")
print("Log-Rank pvalue and HR for Resistant Pancreatic Cancer")
cat(rep("-", 40), "\n")
for (col in names(km_models_resistant)) {
  logrank_p_resistant <- log_rank_results_df_resistant$logrank_p_value[log_rank_results_df_resistant$variable == col]
  cox_HR_resistant <- coxph_df_resistant$cox_hazard_ratio[coxph_df_resistant$column_names == col]
  cat("Variable:", col, "\n")
  cat("Log-Rank P-value:", logrank_p_resistant, "\n")
  cat("Hazard Ratio:", cox_HR_resistant, "\n")
}



```

```{r}
# Define the variables of interest for the resistant dataset
variables_of_interest <- c("Albumin_3.5", "LMR_2.1", "MON_0.8", "LY_1.5", "ANC_4", "ANC_8", 
                                     "NLR_3", "NLR_5", "PLT_300", "Alk_Phos_135", "Alk_Phos_200", 
                                     "Age_60", "Age_65", "Age_70")

```



```{r NPT KM Plot }
# Open a PDF device
pdf("output/Pancreatic_Output/npt_km_plots.pdf")

# Loop through each variable and save its plot on a separate page
for (variable in variables_of_interest) {
  legend_label <- legend_labels_npt[[variable]]
  logrank_p <- log_rank_results_df_npt$logrank_p_value[log_rank_results_df_npt$variable == variable]
  cox_HR <- coxph_df_npt$cox_hazard_ratio[coxph_df_npt$column_names == variable]
  
  # Extract the variable values from the dataframe
  variable_values <- npt_df[[variable]]
  
  # Make sure the event status is logical
  npt_df$Event_Status <- as.logical(npt_df$Event_Status)
  
  # Create the survival object
  surv_obj <- Surv(time = npt_df$time_diff_months, event = npt_df$Event_Status)
  
  # Fit Kaplan-Meier model
  km_fit <- survfit(surv_obj ~ variable_values, data = npt_df)
  
  # Convert p-value and hazard ratio to scientific notation with 3 significant figures
  logrank_p <- format(logrank_p, scientific = TRUE, digits = 3)
  cox_HR <- format(cox_HR, scientific = TRUE, digits = 3)
  
  # Create Kaplan-Meier plot
  km_plot <- ggsurvplot(
    km_fit,
    data = npt_df,
    title = paste("Kaplan-Meier Curve of NPT- Pancreatic Cancer by", variable),
    censor = TRUE,
    xlab = "Time (Months)",
    ylab = "Survival Probability",
    conf.int = TRUE,
    conf.int.style = "step",
    conf.int.alpha = 0.2,
    surv.median.line = "hv",
    xlim = c(0, 24),
    break.time.by = 3,
    breaks = seq(0, 24, by = 3),
    surv.scale = "percent",
    legend.labs = c(legend_label[1], legend_label[2]),  # Assuming two groups for now
    palette = "lancet"
  )
  
  # Add annotation to the plot
  km_plot <- km_plot$plot + annotate(
    "text", x = 0, y = 0.05, 
    label = paste("Log-Rank p-value:", logrank_p, "\nHazard Ratio:", cox_HR), 
    hjust = 0, vjust = 0
  )
  
  # Apply custom theme
  km_plot <- km_plot + ggsurvfit::theme_ggsurvfit_KMunicate()
  
  # Save the plot on a separate page
  print(km_plot)
}

# Close the PDF device
dev.off()


```
```{r Resistant KM Plots}
# Open a PDF device
pdf("output/Pancreatic_Output/resistant_km_plots.pdf")

# Define the variables of interest for the resistant dataset
variables_of_interest_resistant <- c("Albumin_3.5", "LMR_2.1", "MON_0.8", "LY_1.5", "ANC_4", "ANC_8", 
                                     "NLR_3", "NLR_5", "PLT_300", "Alk_Phos_135", "Alk_Phos_200", 
                                     "Age_60", "Age_65", "Age_70")

for (variable_resistant in variables_of_interest_resistant) {
  legend_label_resistant <- legend_labels_resistant[[variable_resistant]]
  logrank_p_resistant <- log_rank_results_df_resistant$logrank_p_value[log_rank_results_df_resistant$variable == variable_resistant]
  cox_HR_resistant <- coxph_df_resistant$cox_hazard_ratio[coxph_df_resistant$column_names == variable_resistant]
  
  # Extract the variable values from the resistant dataframe
  variable_values_resistant <- resistant_df[[variable_resistant]]
  
  # Make sure the event status is logical
  resistant_df$Event_Status <- as.logical(resistant_df$Event_Status)
  
  # Create the survival object for the resistant dataset
  surv_obj_resistant <- Surv(time = resistant_df$time_diff_months, event = resistant_df$Event_Status)
  
  # Fit Kaplan-Meier model for the resistant dataset
  km_fit_resistant <- survfit(surv_obj_resistant ~ variable_values_resistant, data = resistant_df)
  
  # Convert p-value and hazard ratio to scientific notation with 3 significant figures
  logrank_p_resistant <- format(logrank_p_resistant, scientific = TRUE, digits = 3)
  cox_HR_resistant <- format(cox_HR_resistant, scientific = TRUE, digits = 3)
  
  # Create Kaplan-Meier plot for the resistant dataset
  km_plot_resistant <- ggsurvplot(
    km_fit_resistant,
    data = resistant_df,
    title = paste("Kaplan-Meier Curve of Resistant Pancreatic Cancer by", variable_resistant),
    censor = TRUE,
    xlab = "Time (Months)",
    ylab = "Survival Probability",
    conf.int = TRUE,
    conf.int.style = "step",
    conf.int.alpha = 0.2,
    surv.median.line = "hv",
    xlim = c(0, 24),
    break.time.by = 3,
    breaks = seq(0, 24, by = 3),
    surv.scale = "percent",
    legend.labs = legend_label_resistant,  # Assuming two groups for now
    palette = "lancet"
  )
  
  # Add annotation to the resistant dataset plot
  km_plot_resistant <- km_plot_resistant$plot + annotate(
    "text", x = 0, y = 0.05, 
    label = paste("Log-Rank p-value:", logrank_p_resistant, "\nHazard Ratio:", cox_HR_resistant), 
    hjust = 0, vjust = 0
  )
  
  # Apply custom theme
  km_plot_resistant <- km_plot_resistant + ggsurvfit::theme_ggsurvfit_KMunicate()
  
  # Display the plot for the resistant dataset
  print(km_plot_resistant)
}
# Close the PDF device
dev.off()


```


```{r plot ALAN Score NPT}
# Define the variable of interest
variable <- "Prognostic_Score_ALAN_category"

# Extract legend labels for the variable
legend_label <- legend_labels_npt[[variable]]

# Extract Log-Rank p-value and Hazard Ratio for the variable
logrank_p <- log_rank_results_df_npt$logrank_p_value[log_rank_results_df_npt$variable == variable]
cox_HR <- coxph_df_npt$cox_hazard_ratio[coxph_df_npt$column_names == variable]

# Extract variable values from the dataframe
variable_values <- npt_df[[variable]]

# Create the survival object
surv_obj <- Surv(time = npt_df$time_diff_months, event = npt_df$Event_Status)

# Fit Kaplan-Meier model
km_fit <- survfit(surv_obj ~ variable_values, data = npt_df)

# Convert p-value and hazard ratio to scientific notation with 3 significant figures
logrank_p <- format(logrank_p, scientific = TRUE, digits = 3)
cox_HR <- format(cox_HR, scientific = TRUE, digits = 3)

# Create Kaplan-Meier plot
km_plot <- ggsurvplot(
  km_fit,
  data = npt_df,
  title = paste("Kaplan-Meier Curve of NPT- Pancreatic Cancer by ALAN Score"),
  censor = TRUE,
  xlab = "Time (Months)",
  ylab = "Survival Probability",
  conf.int = TRUE,
  conf.int.style = "step",
  conf.int.alpha = 0.2,
  surv.median.line = "hv",
  xlim = c(0, 24),
  break.time.by = 3,
  breaks = seq(0, 24, by = 3),
  surv.scale = "percent",
  legend.labs = legend_label,  # Assuming three groups for "Prognostic_Score_ALAN_category"
  palette = "lancet"
)

# Add annotation to the plot
km_plot <- km_plot$plot + annotate(
  "text", x = 0, y = 0.05, 
  label = paste("Log-Rank p-value:", logrank_p, "\nHazard Ratio:", cox_HR), 
  hjust = 0, vjust = 0
)

# Apply custom theme
km_plot <- km_plot + ggsurvfit::theme_ggsurvfit_KMunicate()

# Display the plot
print(km_plot)


ggsave(filename = "output/Pancreatic_Output/KM Plots/npt_pancreatic_ALAN_kmplot.png", plot = km_plot, width = 10, height = 6)

```
```{r plot ALAN score resistant}
# Define the variable of interest for the resistant dataset
variable_resistant <- "Prognostic_Score_ALAN_category"

# Extract legend labels for the variable
legend_label_resistant <- legend_labels_resistant[[variable_resistant]]

# Extract Log-Rank p-value and Hazard Ratio for the variable
logrank_p_resistant <- log_rank_results_df_resistant$logrank_p_value[log_rank_results_df_resistant$variable == variable_resistant]
cox_HR_resistant <- coxph_df_resistant$cox_hazard_ratio[coxph_df_resistant$column_names == variable_resistant]

# Extract variable values from the resistant dataframe
variable_values_resistant <- resistant_df[[variable_resistant]]

# Create the survival object for the resistant dataset
surv_obj_resistant <- Surv(time = resistant_df$time_diff_months, event = resistant_df$Event_Status)

# Fit Kaplan-Meier model for the resistant dataset
km_fit_resistant <- survfit(surv_obj_resistant ~ variable_values_resistant, data = resistant_df)

# Convert p-value and hazard ratio to scientific notation with 3 significant figures
logrank_p_resistant <- format(logrank_p_resistant, scientific = TRUE, digits = 3)
cox_HR_resistant <- format(cox_HR_resistant, scientific = TRUE, digits = 3)

# Create Kaplan-Meier plot for the resistant dataset
km_plot_resistant <- ggsurvplot(
  km_fit_resistant,
  data = resistant_df,
  title = paste("Kaplan-Meier Curve of Resistant Pancreatic Cancer by ALAN Score"),
  censor = TRUE,
  xlab = "Time (Months)",
  ylab = "Survival Probability",
  conf.int = TRUE,
  conf.int.style = "step",
  conf.int.alpha = 0.2,
  surv.median.line = "hv",
  xlim = c(0, 24),
  break.time.by = 3,
  breaks = seq(0, 24, by = 3),
  surv.scale = "percent",
  legend.labs = legend_label_resistant,  # Assuming three groups for "Prognostic_Score_ALAN_category"
  palette = "lancet"
)

# Add annotation to the resistant dataset plot
km_plot_resistant <- km_plot_resistant$plot + annotate(
  "text", x = 0, y = 0.05, 
  label = paste("Log-Rank p-value:", logrank_p_resistant, "\nHazard Ratio:", cox_HR_resistant), 
  hjust = 0, vjust = 0
)

# Apply custom theme
km_plot_resistant <- km_plot_resistant + ggsurvfit::theme_ggsurvfit_KMunicate()

# Display the plot for the resistant dataset
print(km_plot_resistant)

ggsave(filename = "output/Pancreatic_Output/KM Plots/resistant_pancreatic_ALAN_kmplot.png", plot = km_plot_resistant, width = 10, height = 6)

```