CytoProfile
The goal of CytoProfile is to conduct quality control using biological meaningful cutoff on raw measured values of cytokines. Specifically, test on distributional symmetry to suggest the adopt of transformation. Conduct exploratory analysis including summary statistics, generate enriched barplots, and boxplots. Further, conduct univariate analysis and multivariate analysis for advance analysis.
Installation
Before installation of the CytoProfile package, make sure to install BiocManager and mix0mics packages using:
## install BiocManager
if (!requireNamespace("BiocManager", quietly = TRUE)) install.packages("BiocManager")
## install mixOmics
BiocManager::install('mixOmics')You can install the development version of CytoProfile from GitHub with:
# install.packages("devtools")
devtools::install_github("saraswatsh/CytoProfile")Install CytoProfile from CRAN with:
install.packages("CytoProfile")Example
Below are examples of using the functions provided in CytoProfile. Any saved or generated files that are PDF or PNG format will be found at in the Output Folder.
1. Data Loading and set up
# Loading all packages required
# Data manipulation and reshaping
library(dplyr) # For data filtering, grouping, and summarising.
library(tidyr) # For reshaping data (e.g., pivot_longer, pivot_wider).
# Plotting and visualization
library(ggplot2) # For creating all the ggplot-based visualizations.
library(gridExtra) # For arranging multiple plots on a single page.
library(ggrepel) # For improved label placement in plots (e.g., volcano plots).
library(gplots) # For heatmap.2, which is used to generate heatmaps.
library(plot3D) # For creating 3D scatter plots in PCA and sPLS-DA analyses.
library(reshape2) # For data transformation (e.g., melt) in cross-validation plots.
# Statistical analysis
library(mixOmics) # For multivariate analyses (PCA, sPLS-DA, etc.).
library(e1071) # For computing skewness and kurtosis.
library(pROC) # For ROC curve generation in machine learning model evaluation.
# Machine learning
library(xgboost) # For building XGBoost classification models.
library(randomForest) # For building Random Forest classification models.
library(caret) # For cross-validation and other machine learning utilities.
# Package development and document rendering
library(knitr) # For knitting RMarkdown files and setting chunk options.
library(devtools) # For installing the development version of the package from GitHub.
# Load in the CytoProfile package
library(CytoProfile)
# Loading in data
data("ExampleData1")
data_df <- ExampleData1
# Creating a temporary directory to store output files
path <- tempdir(check = TRUE)2. Exploratory Data Analysis
Boxplots
# Generating boxplots to check for outliers for raw values
cyt_bp(data_df[, -c(1:3)],
pdf_title = file.path(path, "boxplot_by_cytokine_raw.pdf"))
# Removing the first 3 columns to retain only continuous variables.
# Generating boxplots to check for outliers for log2 values
cyt_bp(data_df[, -c(1:3)],
pdf_title = file.path(path, "boxplot_by_cytokine_log2.pdf"),
scale = "log2") Group-Specific Boxplots
# Raw values for group-specific boxplots
cyt_bp2(data_df[, -c(3)],
pdf_title = file.path(path, "boxplot_by_group_and_treatment_raw.pdf"),
scale = NULL)
# Log2-transformed group-specific boxplots
cyt_bp2(data_df[, -c(3)],
pdf_title = file.path(path, "boxplot_by_group_and_treatment_log2.pdf"),
scale = "log2")4. Error Bar Plots
Basic Error Bar Plot
cytokine_mat <- ExampleData1[, -c(1:3)]
cytokineNames <- colnames(cytokine_mat)
nCytokine <- length(cytokineNames)
results <- cyt_skku(ExampleData1[, -c(3)], print_res_log = TRUE,
group_cols = c("Group", "Treatment"))
pdf(file = file.path(path, "bar_error_plot.pdf"))
oldpar <- par(no.readonly = TRUE)
par(mfrow = c(2,2), mar = c(4, 4, 2, 1))
for (k in 1:nCytokine) {
center_df <- data.frame(name = rownames(results[,,k]), results[,,k])
cyt_errbp(center_df,
p_lab = FALSE, es_lab = FALSE, class_symbol = TRUE,
y_lab = "Concentration in log2 scale", main = cytokineNames[k])
}
par(oldpar)
if (dev.cur() > 1) dev.off()
#> agg_png
#> 2Enriched Error Bar Plot with p-values and Effect Sizes
data_df <- ExampleData1[, -c(3)]
cyt_mat <- log2(data_df[, -c(1:2)])
data_df1 <- data.frame(data_df[, 1:2], cyt_mat)
cytokineNames <- colnames(cyt_mat)
nCytokine <- length(cytokineNames)
condt <- !is.na(cyt_mat) & (cyt_mat > 0)
Cutoff <- min(cyt_mat[condt], na.rm = TRUE) / 10
p_aov_mat <- matrix(NA, nrow = nCytokine, ncol = 3)
dimnames(p_aov_mat) <- list(cytokineNames, c("Group", "Treatment", "Interaction"))
p_groupComp_mat <- matrix(NA, nrow = nCytokine, ncol = 3)
dimnames(p_groupComp_mat) <- list(cytokineNames, c("2-1", "3-1", "3-2"))
ssmd_groupComp_stm_mat <- mD_groupComp_stm_mat <- p_groupComp_stm_mat <- p_groupComp_mat
for (i in 1:nCytokine) {
Cytokine <- (cyt_mat[, i] + Cutoff)
cytokine_aov <- aov(Cytokine ~ Group * Treatment, data = data_df)
aov_table <- summary(cytokine_aov)[[1]]
p_aov_mat[i, ] <- aov_table[1:3, 5]
p_groupComp_mat[i, ] <- TukeyHSD(cytokine_aov)$Group[1:3, 4]
p_groupComp_stm_mat[i, ] <- TukeyHSD(cytokine_aov)$`Group:Treatment`[1:3, 4]
mD_groupComp_stm_mat[i, ] <- TukeyHSD(cytokine_aov)$`Group:Treatment`[1:3, 1]
ssmd_groupComp_stm_mat[i, ] <- mD_groupComp_stm_mat[i, ] / sqrt(2 * aov_table["Residuals", "Mean Sq"])
}
results <- cyt_skku(ExampleData1[, -c(3)], print_res_log = TRUE, group_cols = c("Group", "Treatment"))
pdf(file = file.path(path, "bar_error_plot_enriched.pdf"))
oldpar <- par(no.readonly = TRUE)
par(mfrow = c(2,2), mar = c(3, 3, 2, 1))
for (k in 1:nCytokine) {
result_mat <- results[1:9, , k]
center_df <- data.frame(
name = rownames(result_mat),
result_mat[, c("center", "spread")],
p.value = c(1, p_groupComp_stm_mat[k, 1:2]),
effect.size = c(0, ssmd_groupComp_stm_mat[k, 1:2])
)
cyt_errbp(center_df, p_lab = TRUE, es_lab = TRUE,
class_symbol = TRUE, y_lab = "Concentration in log2 scale",
main = cytokineNames[k])
}
par(oldpar)
if (dev.cur() > 1) dev.off()
#> agg_png
#> 25. Univariate Analysis
Two Sample T-test and Mann Whitney U Test
# Performing Two Sample T-test and Mann Whitney U Test
data_df <- ExampleData1[, -c(3)]
data_df <- dplyr::filter(data_df, Group != "ND", Treatment != "Unstimulated")
# Two sample T-test
cyt_ttest(data_df[, c(1:2, 5:6)], scale = "log2", verbose = TRUE, format_output = TRUE)
#> Outcome Categorical Comparison P_value
#> 1 IFN.G Group PreT2D vs T2D 0.0208
#> 2 IL.10 Group PreT2D vs T2D 0.0248
#> 3 IFN.G Treatment CD3/CD28 vs LPS 0.0000
#> 4 IL.10 Treatment CD3/CD28 vs LPS 0.0001
# Mann-Whitney U Test
cyt_ttest(data_df[, c(1:2, 5:6)], verbose = TRUE)
#> $IFN.G_Group
#> [1] 0.0085
#>
#> $IL.10_Group
#> [1] 0.0119
#>
#> $IFN.G_Treatment
#> [1] 0
#>
#> $IL.10_Treatment
#> [1] 06. Multivariate Analysis
Partial Least Squares Discriminant Analysis (PLS-DA)
# cyt_plsda function.
data <- ExampleData1[, -c(3)]
data_df <- dplyr::filter(data, Group != "ND" & Treatment == "CD3/CD28")
cyt_splsda(data_df,
pdf_title = file.path(path, "example_spls_da_analysis.pdf"),
colors = c("black", "purple"),
bg = FALSE, scale = "log2", ellipse = TRUE,
conf_mat = FALSE, var_num = 25,
cv_opt = "loocv", comp_num = 2,
pch_values = c(16, 4), group_col = "Group", group_col2 = "Treatment",
roc = TRUE)
#> agg_png
#> 27. Principal Component Analysis (PCA)
data <- ExampleData1[, -c(3,23)]
data_df <- dplyr::filter(data, Group != "ND" & Treatment != "Unstimulated")
cyt_pca(data_df,
pdf_title = file.path(path, "example_pca_analysis.pdf"),
colors = c("black", "red2"), scale = "log2",
comp_num = 3, pch_values = c(16, 4),
style = "3D", group_col = "Group", group_col2 = "Treatment")
#> [1] "Results based on log2 transformation:"
#> agg_png
#> 2
cyt_pca(data_df,
pdf_title = file.path(path, "example_pca_analysis_2.pdf"),
colors = c("black", "red2"), scale = "log2",
comp_num = 2, pch_values = c(16, 4), group_col = "Group")
#> [1] "Results based on log2 transformation:"
#> agg_png
#> 29. Heatmap
# Generating Heat map
cyt_heatmap(data = data_df,
scale = "log2", # Optional scaling
annotation_col_name = "Group",
title = NULL)10. Dual Flashlight Plot
# Generating dual flashlights plot
data_df <- ExampleData1[, -c(2:3)]
dfp <- cyt_dualflashplot(data_df, group_var = "Group",
group1 = "T2D", group2 = "ND",
ssmd_thresh = -0.2, log2fc_thresh = 1,
top_labels = 10)
# Print the plot
dfp
# Print the table data used for plotting
print(dfp$data)
#> # A tibble: 25 × 11
#> cytokine mean_ND mean_PreT2D mean_T2D variance_ND variance_PreT2D
#> <chr> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 CCL.20.MIP.3A 634. 404. 887. 6.72e+ 5 2.74e+5
#> 2 GM.CSF 2.65 3.11 1.92 2.63e+ 1 3.14e+1
#> 3 IFN.G 57730. 18303. 61484. 2.86e+10 2.30e+9
#> 4 IL.10 979. 836. 1366. 1.99e+ 6 1.19e+6
#> 5 IL.12.P70 13.0 39.1 78.9 4.15e+ 2 2.56e+4
#> 6 IL.13 1064. 1543. 1122. 5.60e+ 6 1.11e+7
#> 7 IL.15 7.92 4.29 8.22 3.54e+ 1 2.58e+1
#> 8 IL.17A 352. 653. 615. 9.40e+ 5 2.88e+6
#> 9 IL.17E.IL.25 0.0101 0.0163 0.01 1.01e- 6 3.88e-3
#> 10 IL.17F 1.63 2.35 3.11 1.56e+ 1 3.37e+1
#> # ℹ 15 more rows
#> # ℹ 5 more variables: variance_T2D <dbl>, ssmd <dbl>, log2FC <dbl>,
#> # SSMD_Category <chr>, Significant <lgl>11. Machine Learning Models
Using XGBoost for classification
# Using XGBoost for classification
data_df0 <- ExampleData1
data_df <- data.frame(data_df0[, 1:3], log2(data_df0[, -c(1:3)]))
data_df <- data_df[, -c(2:3)]
data_df <- dplyr::filter(data_df, Group != "ND")
cyt_xgb(data = data_df, group_col = "Group",
nrounds = 500, max_depth = 4, eta = 0.05,
nfold = 5, cv = TRUE, eval_metric = "mlogloss",
early_stopping_rounds = NULL, top_n_features = 10,
verbose = 0, plot_roc = TRUE, print_results = FALSE)Using Random Forest for classification
# Using Random Forest for classification
cyt_rf(data = data_df, group_col = "Group", k_folds = 5,
ntree = 1000, mtry = 4, run_rfcv = TRUE,
plot_roc = TRUE, verbose = FALSE)