In 2021 and 2022, 18 of 31 spider mite populations tested from Ontario soybean fields were classified as resistant to dimethoate. On the plants, mite feeding can leave tiny yellow spots on the leaves (Ontario Grain Farmer, 2026). What changes inside a plant as mites continue to feed?
The two-spotted spider mite (Tetranychus urticae), the herbivore examined in this project.
Image credit. CSIRO (2000), Wikimedia Commons, CC BY 3.0.
Host plants can influence how mites respond to pesticides. Dermauw et al. (2013) found that two-spotted spider mites (Tetranychus urticae) adapted to tomato were less susceptible to several pesticides. The Grbić lab and its collaborators have also examined how plants respond to mite feeding. Zhurov et al. (2014) documented changes in gene expression in Arabidopsis thaliana after mite feeding, and Díaz-Riquelme et al. (2016) identified changes in defence-related gene expression in grapevine.
Studies in other plants link mite feeding to the flavonoid pathway. Li et al. (2025) reported that mite feeding stimulated flavonoid biosynthesis in lima bean. Chen et al. (2022) found that mite infestation increased the expression of chalcone synthase (CHS), a gene involved in flavonoid biosynthesis, in cassava. These findings motivate the question for this project. Does mean CHS expression in Arabidopsis differ among an untreated control and five mite-feeding durations?
This project applies quantitative methods to study plant responses to herbivory, or feeding by a plant-eating organism. This is a research area of the Grbić lab. I use a one-way analysis of variance (ANOVA) in R to compare mean CHS expression across six experimental conditions in Kon, a naturally occurring line of Arabidopsis thaliana. I plot the individual measurements and group means and inspect the model residuals to assess the assumptions.
The data come from NCBI Gene Expression Omnibus series GSE49981,
deposited with the study by Zhurov et al. (2014). The study used an
Affymetrix ATH1 microarray to estimate the amount of RNA from each gene.
The CHS Expression column in Geness.csv
contains the published readings for the CHS probe set
AT5G13930_at, processed using robust multiarray average
(RMA) so the readings can be compared across samples. This analysis uses
the Kon feeding-duration samples and untreated controls. NCBI
GEO series GSE49981
The six groups are an untreated control and groups exposed to 10 two-spotted spider mites per plant for 1, 3, 6, 12, or 24 hours. GEO labels the untreated samples as 24-hour controls. Each group has two biological replicates, and each replicate comprises pooled tissue from six plants. Each CSV row therefore represents one pooled biological replicate, giving 12 observations for the analysis.
I rename the CHS expression column for use in R and set
Treatment as a categorical factor ordered from the
untreated control through the five feeding durations. I inspect the data
structure before fitting the model.
library(readr)
library(dplyr)
##
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
##
## filter, lag
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
library(ggplot2)
library(ggfortify)
treatment_levels <- c("Control", "1 h", "3 h", "6 h", "12 h", "24 h")
chs <- read_csv("D:/Ph.d/course Fall 1/BiostatisticR/Notebook/Session3/Geness.csv") %>%
rename(CHS_RMA_Expression = `CHS Expression`) %>%
mutate(Treatment = factor(Treatment, levels = treatment_levels))
## Rows: 12 Columns: 8
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (4): GEO-Sample-ID, GEO-Sample-Title, Treatment, Biological Replicate
## dbl (4): Feeding-Hours, Mites Per Plant, Plants Per Replicate, CHS Expression
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
glimpse(chs)
## Rows: 12
## Columns: 8
## $ `GEO-Sample-ID` <chr> "GSM1211076", "GSM1211077", "GSM1211078", "GSM1…
## $ `GEO-Sample-Title` <chr> "KON_CNTRL_24H_A", "KON_CNTRL_24H_B", "KON_01H_…
## $ Treatment <fct> Control, Control, 1 h, 1 h, 3 h, 3 h, 6 h, 6 h,…
## $ `Feeding-Hours` <dbl> 0, 0, 1, 1, 3, 3, 6, 6, 12, 12, 24, 24
## $ `Mites Per Plant` <dbl> 0, 0, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10
## $ `Biological Replicate` <chr> "A", "B", "A", "B", "A", "B", "A", "B", "A", "B…
## $ `Plants Per Replicate` <dbl> 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6
## $ CHS_RMA_Expression <dbl> 7.722787, 7.844867, 8.022964, 8.392698, 9.38507…
I use a horizontal boxplot to explore the centre and spread of CHS expression across the six conditions. Each group contains two pooled biological replicates.
ggplot(chs, aes(x = Treatment, y = CHS_RMA_Expression)) +
geom_boxplot() +
theme_bw()
Vertical boxplots of normalized CHS expression across the six mite-feeding conditions.
ggplot(chs, aes(x = Treatment, y = CHS_RMA_Expression)) +
geom_boxplot() +
theme_bw() +
coord_flip()
Horizontal boxplots of normalized CHS expression across the six mite-feeding conditions.
I use a one-way ANOVA to test whether mean CHS expression differs among the six mite-feeding conditions.
model_gene <- lm(CHS_RMA_Expression ~ Treatment, data = chs)
autoplot(model_gene, smooth.linetype = "blank")
## Warning: `fortify(<lm>)` was deprecated in ggplot2 4.0.0.
## ℹ Please use `broom::augment(<lm>)` instead.
## ℹ The deprecated feature was likely used in the ggfortify package.
## Please report the issue at <https://github.com/sinhrks/ggfortify/issues>.
## This warning is displayed once per session.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
## Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
## ℹ Please use `linewidth` instead.
## ℹ The deprecated feature was likely used in the ggfortify package.
## Please report the issue at <https://github.com/sinhrks/ggfortify/issues>.
## This warning is displayed once per session.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
anova(model_gene)
## Analysis of Variance Table
##
## Response: CHS_RMA_Expression
## Df Sum Sq Mean Sq F value Pr(>F)
## Treatment 5 3.7423 0.74847 20.15 0.001092 **
## Residuals 6 0.2229 0.03714
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
The one-way ANOVA provided evidence that at least one of the six group means differed (F(5, 6) = 20.15, p = 0.00109).
model_gene <- lm(CHS_RMA_Expression ~ Treatment, data = chs)
summary(model_gene)
##
## Call:
## lm(formula = CHS_RMA_Expression ~ Treatment, data = chs)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.2192 -0.0831 0.0000 0.0831 0.2192
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.7838 0.1363 57.116 1.93e-09 ***
## Treatment1 h 0.4240 0.1927 2.200 0.070102 .
## Treatment3 h 1.4520 0.1927 7.534 0.000283 ***
## Treatment6 h 1.5377 0.1927 7.978 0.000207 ***
## Treatment12 h 1.1432 0.1927 5.931 0.001024 **
## Treatment24 h 0.6325 0.1927 3.282 0.016784 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.1927 on 6 degrees of freedom
## Multiple R-squared: 0.9438, Adjusted R-squared: 0.897
## F-statistic: 20.15 on 5 and 6 DF, p-value: 0.001092
I compare each group mean with the control mean.
sumDat <- chs %>%
group_by(Treatment) %>%
summarise(meanCHS = mean(CHS_RMA_Expression), .groups = "drop")
control_mean <- sumDat$meanCHS[sumDat$Treatment == "Control"]
ggplot(chs, aes(x = CHS_RMA_Expression, y = Treatment, colour = Treatment)) +
geom_vline(xintercept = control_mean, linetype = "dashed", colour = "grey40") +
geom_segment(
data = sumDat, aes(x = meanCHS, xend = control_mean, y = Treatment, yend = Treatment),
inherit.aes = FALSE, colour = "black"
) +
geom_point(alpha = 0.4, size = 2) +
geom_point(data = sumDat, aes(x = meanCHS), shape = 18, size = 5) +
scale_y_discrete(limits = rev(treatment_levels)) +
labs(x = "RMA-normalized CHS expression", y = "Mite-feeding condition",
colour = "Treatment") +
theme_bw()
Small points are pooled biological replicates. Diamonds show group means. The dashed line marks the control mean, and horizontal segments show observed differences from that reference.
The 6-hour group had the highest observed mean CHS expression, followed by the 3-hour group.
sumDat <- chs %>%
group_by(Treatment) %>%
summarise(meanCHS = mean(CHS_RMA_Expression))
ggplot(chs, aes(x = CHS_RMA_Expression, y = Treatment, colour = Treatment)) +
geom_point(alpha = 0.4, size = 2) +
geom_point(data = sumDat, aes(x = meanCHS), shape = 18, size = 5) +
scale_y_discrete(limits = treatment_levels) +
scale_colour_manual(values = c(
"Control" = "red",
"1 h" = "green4",
"3 h" = "turquoise3",
"6 h" = "purple",
"12 h" = "orange3",
"24 h" = "steelblue"
)) +
labs(
x = "RMA-normalized CHS expression",
y = "Mite-feeding condition",
colour = "Treatment"
) +
theme_bw()
Small points show individual pooled biological replicates, and diamonds show the mean CHS expression for each group.
Gene expression. Here, the amount of RNA produced
from a gene, estimated from the microarray measurements.
Chalcone synthase (CHS). An enzyme plants use to make
flavonoids. In this report, CHS expression refers to the amount of RNA
from the gene that encodes this enzyme. Flavonoids. A
group of compounds made by plants. Some contribute to plant responses to
herbivore feeding. RMA normalization. A method for
processing microarray readings so expression can be compared across
samples. Pooled biological replicate. One sample formed
by combining tissue from six plants. Each pooled sample provides one CHS
expression measurement and occupies one row in Geness.csv.
One-way ANOVA. A statistical test that compares group
means for one factor. Here, that factor is the mite-feeding condition.
Residual. The difference between an observed CHS
expression value and the value predicted by the model for its group.
Normal Q–Q plot. A graph used to assess whether the
model residuals approximately follow a normal distribution.