2. Evaluating the in vitro tumorigenicity of CSC candidate using the 3D multi-spheroid model by determining object area using 3D-SiSP analytical method
Krittiyabhorn Kongtanawanich
2025-02-25
Overview of R analysis for evaluating the in vitro tumorigenicity of CSC candidate using the 3D multi-spheroid model by determining object area using 3D-SiSP analytical method.
The in vitro tumorigenic assay is a crucial step in verifying CSC
properties. In our study, we focus on the usage of the CSC biosensor
SORE6, however, other CSC markers can also be utilized with our analysis
platform. SORE6 serves as the CSC biosensor in this example, with a
green fluorescence protein (GFP) acting as the biosensor protein in this
system. Cells containing the SORE6 biosensor were sorted into
GFP-positive (POS) and GFP-negative (NEG)
populations using fluorescence-activated cell sorting, expected to
represent CSCs and non-CSCs, respectively. Then, the expected CSC and
non-CSC populations were separately cultured in the 3D multi-spheroid
model at various initial seeding densities. According to the 3D-SiSP
analytical method, the tumorigenicity of CSC and non-CSC were compared
by object area.
According to image analysis, all cells in each well were nuclear stained, enabling the measurement of the area of all objects within the well (Figure 1). To evaluate the in vitro tumorigenicity of cancer stem cell (CSC) candidates, we developed a high-throughput analysis to determine objects area using 3D Surface Integrative Spheroid Profiling (3D-SiSP) analytical method.
In our analysis pipeline, the area of all objects was included without cut-off concern. Subsequently, the objects in each initial cell seeding condition were used for object area calculation and compared between the GFP-positive and GFP-negative groups. Statistical tests were performed by comparing the CSC and non-CSC groups in each cell seeding density. Additionally, box plots were generated for data visualization.
Figure 1: Image analysis of each 3D object. (Created by BioRender.com /
Mahidol University)
1. Analysis overview
Here, the analysis composed of 4 main sections below.
- Section 1: Define information of the dataset.
- Section 2: Create analysis pipeline.
- Part 2.1: Create a function (
Object.Area) for mean calculation of area of all objects (ObjectArea).
- Part 2.2: Perform the analysis for all conditions using the created
function (
Object.Area).
- Part 2.3: Create summary analysis
- Part 2.1: Create a function (
- Section 3: Statistic test
- Part 3.1: Normality test
- Part 3.2: Statistical test
- Part 3.1: Normality test
- Section 4: Data visualization.
2. Inputs
The plate map was arranged as shown in the picture below. Biosensor
groups (POS and NEG) and initial cell seeding
density were arranged accordingly. According to the image analysis, the
area of all objects were measured and stored in a
WellName.csv file, well by well.
In the WellName.csv files contains 4 columns, which are
described as below.
Object_Area_um2= Area of each object (μm²)
CellSeedingDensity= Initial cell seeding density
BiosensorGroup= CSC biosensor groups (POSorNEG)
Replicate= Technical replicate of the experiment
Here, we also provide a
set of example inputs, where each WellName.csv file contains
information about the objects in the respective well.
Figure 2: Plate map of in vitro tumorigenic assay. (Created by
BioRender.com / Mahidol University)
Analysis for evaluating the in vitro tumorigenicity of CSC candidate using the 3D multi-spheroid model by determining object area using 3D-SiSP analytical method.
Import R library
library(dplyr)
library(ggplot2)
library(ggpubr)
library(rstatix)Section 1: Define information of the dataset.
To define information of the dataset.
1.1 Define working directory: Output_Directory
1.2 Define input directory: Input_Directory
1.3 Define the cell line name: CellLine_Name
# Dataset example
## Output directory
Output_Directory <- 'path/to/directory'
setwd(Output_Directory)
## Input directory
Input_Directory <- 'path/to/directory'
CellLine_Name <- "CellLineX"Section 2: Create analysis pipeline.
Part 2.1: Create a function (Object.Area) for mean
calculation of area of all objects (ObjectArea).
Object.Area <- function(Well_FileName, Input_Directory, CellLine_Name)
{
# Import each `.csv` file.
Well_FilePath <- file.path(Input_Directory, Well_FileName)
Well_File <- read.csv(Well_FilePath)
# Extract information (Initial cell seeding density, Biosensor group) from the file.
CellSeedingDensity <- Well_File$CellSeedingDensity[1]
BiosensorGroup <- Well_File$BiosensorGroup[1]
# Extract area of all objects from the file.
EachWell_ObjectArea <- Well_File$Object_Area_um2
EachWell_ObjectArea <- as.vector(EachWell_ObjectArea)
# Calculate mean of area of all objects in the well
Mean.Object_Area <- mean(EachWell_ObjectArea)
# Create a data frame to store object area information.
Result <- data.frame(FileName = Well_FileName,
CellSeedingDensity = CellSeedingDensity,
BiosensorGroup = BiosensorGroup,
CellLine = CellLine_Name,
ObjectArea = Mean.Object_Area
)
return(Result)
}Part 2.2: Perform the analysis for all conditions using the created
function (Object.Area)
# Create a blank data frame to store `EachWell_Object_Area` information.
AllCondition_Object_Area <- data.frame()
# Load the input data.
FileName_list <- list.files(Input_Directory)
FileName_list <- as.list(FileName_list)
for (Well in FileName_list) {
# Calculate the mean of object area in each well by `Object.Area` function.
EachWell_Object_Area <- Object.Area(Well, Input_Directory, CellLine_Name)
# Store `EachWell_Object_Area` into the created data frame (`AllCondition_Object_Area`).
AllCondition_Object_Area <- rbind(AllCondition_Object_Area,
EachWell_Object_Area
)
}
# Reorder the data according to `BiosensorGroup` and `CellSeedingDensity`.
AllCondition_Object_Area <- with(AllCondition_Object_Area,
AllCondition_Object_Area[order(BiosensorGroup, CellSeedingDensity), ]
)
# Write the `AllCondition_Object_Area` data into a `.csv` file.
write.csv(AllCondition_Object_Area,
paste0(CellLine_Name, "_Raw_ObjectArea.csv"), row.names = FALSE)
print(paste0("The .csv file was written as '", CellLine_Name, "_Raw_ObjectArea.csv'"))## [1] "The .csv file was written as 'CellLineX_Raw_ObjectArea.csv'"Observe the AllCondition_Object_Area data.
head(AllCondition_Object_Area)## FileName CellSeedingDensity BiosensorGroup CellLine ObjectArea
## 2 B11.csv 63 NEG CellLineX 850.4693
## 12 C11.csv 63 NEG CellLineX 946.3654
## 22 D11.csv 63 NEG CellLineX 610.4338
## 32 E11.csv 63 NEG CellLineX 556.7345
## 42 F11.csv 63 NEG CellLineX 803.2494
## 52 G11.csv 63 NEG CellLineX 331.7023Part 2.3: Create summary analysis.
For the summary analysis of the spheroid count
(AllCondition_Object_Area), the statistics comprising mean
and standard deviation were calculated and written as a
.csv file.
Summary_ObjectArea <- AllCondition_Object_Area%>%
group_by(CellLine,
BiosensorGroup,
CellSeedingDensity) %>%
summarise_at(vars(ObjectArea),
list(Mean = mean,
SD = sd))
# Write the `Summary_Result` into a `.csv` file.
write.csv(Summary_ObjectArea,
paste0(CellLine_Name, "_Summary_ObjectArea.csv"), row.names = FALSE)
print(paste0("Summary of the object area analysis was written as '", CellLine_Name, "_Summary_ObjectArea.csv'"))## [1] "Summary of the object area analysis was written as 'CellLineX_Summary_ObjectArea.csv'"Observe the Summary_ObjectArea data.
head(Summary_ObjectArea)## # A tibble: 6 × 5
## # Groups: CellLine, BiosensorGroup [2]
## CellLine BiosensorGroup CellSeedingDensity Mean SD
## <chr> <chr> <int> <dbl> <dbl>
## 1 CellLineX NEG 63 683. 226.
## 2 CellLineX NEG 125 715. 164.
## 3 CellLineX NEG 250 801. 94.4
## 4 CellLineX NEG 500 1431. 234.
## 5 CellLineX NEG 1000 2251. 175.
## 6 CellLineX POS 63 742. 239.Section 3: Statistical test
3.1 Normality test
To determine whether the data follows a normal distribution, the normality test should be performed.
shapiro.test(AllCondition_Object_Area$ObjectArea)##
## Shapiro-Wilk normality test
##
## data: AllCondition_Object_Area$ObjectArea
## W = 0.88436, p-value = 3.71e-05For the example dataset, the result showed that the data do not follow a normal distribution.
3.2 Statistical test
According to the non-normal distribution of the data, the Wilcoxon
rank-sum test was used for statistical testing by comparing the spheroid
count between the biosensor groups in the same cell seeding
density.
If your data follows a normal distribution, a parametric test like
the “T-test” should be used instead.
Here, we also provide the choice of statistical tests as shown
below.
# T-test
Stat_ObjectArea_Ttest <- compare_means(ObjectArea ~ BiosensorGroup,
data = AllCondition_Object_Area,
group.by = "CellSeedingDensity",
method = "t.test"
)
# Wilcoxon rank-sum test
Stat_ObjectArea_Wilcox <- compare_means(ObjectArea ~ BiosensorGroup,
data = AllCondition_Object_Area,
group.by = "CellSeedingDensity",
method = "wilcox.test"
)
# Write the `Stat_ObjectArea` into a `.csv` file.
write.csv(Stat_ObjectArea_Wilcox,
paste0(CellLine_Name, "_Stat_ObjectArea.csv"), row.names = FALSE)
print(paste0("Statistical test of the object area was written as '", CellLine_Name, "_Stat_ObjectArea.csv'"))## [1] "Statistical test of the object area was written as 'CellLineX_Stat_ObjectArea.csv'"Observe the Stat_ObjectArea_Wilcox data.
Stat_ObjectArea_Wilcox ## # A tibble: 5 × 9
## CellSeedingDensity .y. group1 group2 p p.adj p.format p.signif method
## <int> <chr> <chr> <chr> <dbl> <dbl> <chr> <chr> <chr>
## 1 63 Objec… NEG POS 0.589 0.59 0.5887 ns Wilco…
## 2 125 Objec… NEG POS 0.00866 0.035 0.0087 ** Wilco…
## 3 250 Objec… NEG POS 0.00866 0.035 0.0087 ** Wilco…
## 4 500 Objec… NEG POS 0.00216 0.011 0.0022 ** Wilco…
## 5 1000 Objec… NEG POS 0.00866 0.035 0.0087 ** Wilco…Section 4: Data visualization
Data visualization is presented as a faceted box plot in an object
named Tumorigenicity_BoxPlot and saved as a
.png file.
# Define `CellSeedingDensity` and `BiosensorGroup` as factors.
AllCondition_Object_Area$CellSeedingDensity <- factor(AllCondition_Object_Area$CellSeedingDensity,
levels = c("1000", "500", "250", "125", "63")
)
AllCondition_Object_Area$BiosensorGroup <- factor(AllCondition_Object_Area$BiosensorGroup,
levels = c("POS", "NEG")
)
# Create a box plot which faceted by `CellSeedingDensity`.
Tumorigenicity_BoxPlot <- ggboxplot(AllCondition_Object_Area,
x = "BiosensorGroup",
y = "ObjectArea",
color = "BiosensorGroup",
fill = "BiosensorGroup",
facet.by = "CellSeedingDensity",
short.panel.labs = TRUE,
add = "jitter") +
scale_color_manual(values = c("#154360",
"#9A7D0A")) +
scale_fill_manual(values = c("#0073C2FF",
"#EFC000FF")) +
labs(x = "GFP expression",
y = "Number of spheroids",
title = "In vitro tumorigenicity") +
theme(plot.title = element_text(hjust = 0.5)) +
guides(fill = guide_legend(title = NULL)) +
facet_wrap(~CellSeedingDensity) +
facet_grid(rows = NULL,
cols = vars(CellSeedingDensity)) +
stat_compare_means(label = "p.format",
label.x.npc = 0.8, label.y.npc =0.9) +
theme(legend.position = "none")
# Save the box plot as a `.png` file.
ggsave(paste0(CellLine_Name, "_Plot_ObjectArea.png"),
width = 10 , height = 3.5,
dpi = 300, units = "in")
print(paste0("The .png file was saved as '", CellLine_Name, "'plot_ObjectArea.png'"))## [1] "The .png file was saved as 'CellLineX'plot_ObjectArea.png'"
Figure 4: In vitro tumorigenic assay of cell line X. The spheroids were
determined by area cut-off 8,000 μm².