Predicting Resistance Trends of ESKAPE Pathogens using R

- 10 mins
machine learning R ESKAPE pathogens public health healthcare epidemiology global health infectious diseases bacteria antibiotic resistance

Bangalore, India 2023

This project is aimed at tackling the escalating threat of antimicrobial resistance (AMR) posed by ESKAPE pathogens. We analyze a the Pfizer ATLAS dataset (8 lakh+ isolates) from the Vivli AMR Register competition (2023) to understand their prevalence, geographical distribution, and resistance trends. Additionally, we leverage machine learning to predict future resistance patterns and the analysis is conducted on a High-Performance Computing (HPC) environment using R Studio to handle the massive dataset size effectively.

Sample figures generated from the project

1. Data Cleaning and Preparation

This code utilizes the dplyr and tidyr libraries to filter and manipulate the raw data based on specific criteria. You can adjust the filter conditions within the filter() function to select desired species, countries, etc. The filtered data is then saved as a new CSV file.

# Sorting the raw data based on requirement
library(dplyr)
library(magrittr)
library(tidyr)

# Read the data from the CSV file
raw_data <- read.csv("/path/to/your/csvfile.csv")

# Filter the data for a specific species and country
# Replace "Species" and "Country" with the appropriate column names
# Replace "Species_Name" and "Country_Name" with the desired species and country names
filtered_data <- raw_data %>%
  filter(Species == "Species_Name" & Country == "Country_Name")

# Save the filtered data to a new CSV file
# Replace "/path/to/your/output/file.csv" with the desired output path and file name
write.csv(filtered_data, '/path/to/your/output/file.csv')

# Remove any empty columns
# Load required libraries
library(dplyr)

# Read the data from the CSV file
# Replace "/path/to/your/input/file.csv" with the actual path to your input CSV file
data <- read.csv("/path/to/your/input/file.csv")

# Identify and remove empty columns
empty_cols <- which(colSums(is.na(data)) == nrow(data))
data <- data[, -empty_cols]

# Save the modified data to a new CSV file
# Replace "/path/to/your/output/modified_file.csv" with the desired output path and file name
write.csv(data, "/path/to/your/output/modified_file.csv", row.names = FALSE)

2. Data Visualisation

# Creating a distribution plot of samples across the world from the provided dataset 
# Load required libraries
library(ggplot2)
library(maps)

# Read the CSV file into a data frame
# Replace "/path/to/your/input/file.csv" with the actual path to your input CSV file
data <- read.csv("/path/to/your/input/file.csv")

# Create a world map
world <- map_data("world")

# Check for missing or non-unique country names in the data
if (any(is.na(data$Country))) {
  stop("There are missing values in the 'Country' column.")
}

if (any(duplicated(data$Country))) {
  stop("There are non-unique values in the 'Country' column.")
}

# Merge the data frame with the world map
data_map <- inner_join(world, data, by = c("region" = "Country"))

# Create the plot using ggplot2
plot <- ggplot() +
  geom_polygon(data = data_map, aes(x = long, y = lat, group = group, fill = SampleID)) +
  coord_map("miller") +
  theme_minimal() +
  theme(axis.text = element_blank(),
        axis.title = element_blank(),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank()) +
  scale_fill_viridis_c() +
  labs(title = "Distribution of Samples on World Map")

# Save the plot as a PNG file
# Replace "/path/to/your/output/sample_map.png" with the desired output path and file name
ggsave("/path/to/your/output/sample_map.png", plot, width = 10, height = 6, dpi = 300)

Analyzing Prevalence by Species and Year: Key Steps

# To generate a plot for prevalence of species by year 
# Load required libraries
library(ggplot2)
library(dplyr)

# Read the CSV file into a data frame
# Replace "/path/to/your/input/file.csv" with the actual path to your input CSV file
data_counts <- read.csv("/path/to/your/input/file.csv")

# Check for missing or non-unique values in the data
if (any(is.na(data_counts$Year)) || any(is.na(data_counts$n))) {
  stop("There are missing values in the 'Year' or 'n' columns.")
}

if (any(duplicated(data_counts$Year))) {
  stop("There are non-unique values in the 'Year' column.")
}

# Create the line plot using ggplot2
line_plot <- ggplot(data_counts, aes(x = Year, y = n, group = Species, color = Species)) +
  geom_line() +
  geom_point() +
  labs(title = "Prevalence of ESKAPE pathogens by species and year",
       x = "Year",
       y = "Number of Samples",
       color = "Species") +
  theme_minimal()

# Display the line plot
print(line_plot)

# Save the plot as a PNG file
# Replace "/path/to/your/output/line_plot.png" with the desired output path and file name
ggsave("/path/to/your/output/line_plot.png", line_plot, width = 10, height = 6, dpi = 300)

3. Analyzing MDR Isolate Count

  1. Data Preprocessing:
# preparing the data for MDR analysis using the AMR package 
# First convert the entries in the dataset columns into R,I,S format from Resistant,Intermediate,Susceptible format 
# Load required libraries
library(dplyr)

# Read the raw data from CSV file
# Replace "/path/to/your/input/raw_data.csv" with the actual path to your input CSV file
raw_data <- read.csv("/path/to/your/input/raw_data.csv")

# Function to replace "Resistant", "Susceptible", and "Intermediate" with "R", "S", and "I" respectively
replace_resistant_susceptible_intermediate <- function(x) {
  x <- gsub("Resistant", "R", x)
  x <- gsub("Susceptible", "S", x)
  gsub("Intermediate", "I", x)
}

# Automatically detect and replace columns containing "Resistant", "Susceptible", and "Intermediate"
raw_data <- raw_data %>%
  mutate(across(where(is.character), replace_resistant_susceptible_intermediate))

# Save the updated data to a new CSV file if needed
# Replace "/path/to/your/output/modified_data.csv" with the desired output path and file name
write.csv(raw_data, "/path/to/your/output/modified_data.csv", row.names = FALSE)



# Determine the MDR isolates count from the prepared dataset usinf AMR package https://msberends.github.io/AMR/reference/mdro.html

mdro(
  # A dataframe antibiotic columns like AMX or amox.
  x = NULL, 
  # A specefic supported National/International/Custom guideline to be followed 
  guideline = "CMI2012",
  # column name of the names or codes of the microorganisms
  col_mo = NULL,
  # a logical to indicate whether progress should be printed to the console
  info = interactive(),
  # minimal required percentage of antimicrobial classes that must be available per isolate
  pct_required_classes = 0.5,
  # a logical to indicate whether all values of S and I must be merged into one, so resistance is only considered when isolates are R, not I. 
  combine_SI = TRUE,
  # a logical to turn Verbose mode on and off (default is off).
  verbose = FALSE,
  # a logical to indicate whether only antibiotic columns must be detected that were transformed to class sir
  only_sir_columns = FALSE,
  ...
)



# Generating a plot for the MDR analysis 
# Load required libraries
library(ggplot2)
library(tidyverse)

# Example data: Replace these with your actual data
pathogen_names <- c("E. faecium", "S. aureus", "K. pneumoniae", "A. baumannii", "P. aeruginosa", "Enterobacter spp")
total_counts <- c(16661, 147921, 88712, 38036, 94460, 48041)
mdr_counts <- c(5710, 1596, 27327, 23077, 17165, 7549)

# Combine vectors into a data frame
mdr_data <- data.frame(pathogen = pathogen_names, total_count = total_counts, mdr_count = mdr_counts)

# Calculate the proportion of MDR isolates relative to total isolates
mdr_data$mdr_proportion <- mdr_data$mdr_count / mdr_data$total_count

# Create the grouped bar plot with text labels
ggplot(mdr_data, aes(x = reorder(pathogen, total_count), y = total_count, fill = "Total")) +
  geom_bar(stat = "identity", position = "dodge") +
  geom_bar(aes(y = mdr_count, fill = "MDR"), stat = "identity", position = "dodge") +
  geom_text(aes(label = paste0(round(mdr_proportion * 100), "%")),
            position = position_dodge(width = 0.9), vjust = -0.5, size = 3.5) +
  labs(title = "MDR Isolate Count Among Total Isolates for ESKAPE Pathogens",
       x = "Pathogen",
       y = "Count",
       fill = "Isolate Type") +
  scale_fill_manual(values = c("Total" = "gray", "MDR" = "red")) +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1),
        legend.title = element_blank(),
        axis.text = element_text(color = "black"))

4. Predicting Future Antibiotic Resistance Trends

This code snippet explores the possibility of predicting future resistance patterns using the AMR package. Here’s a breakdown of the steps involved:

Required Libraries

dplyr for data manipulation tasks. ggplot2 for creating visualizations. AMR for resistance prediction functionalities.

Data Loading:

Resistance Prediction:

Visualizing Predictions:

# Future antimicrobial resistance prediction using regression models 
# AMR package (https://msberends.github.io/AMR/articles/resistance_predict.html)
# Load the necessary libraries
library(dplyr)
library(ggplot2)
library(AMR)

# Read the CSV file into a data frame
# Replace "/path/to/your/input/your_file_path.csv" with the actual path to your input CSV file
data <- read.csv("/path/to/your/input/your_file_path.csv")

# Specify the name of the antibiotic for resistance prediction
# Replace "Your_Antibiotic_Column_Name" with the actual column name containing antibiotic resistance information
antibiotic_column <- "Your_Antibiotic_Column_Name"

# Perform resistance prediction for the specified antibiotic
# Replace "Year" and "Your_Antibiotic_Column_Name" with the actual column names for date and antibiotic resistance information
predict_antibiotic <- data %>%
  resistance_predict(
    col_date = "Year",
    col_ab = antibiotic_column,
    model = "binomial"
  )

# Plot the resistance prediction results
plot(predict_antibiotic)

ggplot_sir_predict(predict_antibiotic)

Related Posts

Srikanth Srinivas

Srikanth Srinivas

Bioinformatics graduate student at University of Bristol | Former Bioinformatics Intern at Global Health Research Unit, CRL, KIMS

rss facebook twitter github youtube mail spotify instagram linkedin google pinterest medium vimeo gscholar