In dwassarman/cellpanelr: Analyze Data from Cell Line Panels

library(cellpanelr)
library(tidyverse) # convenient functions for data joining and manipulation

# Get the nutlin-3 sensitivity data
nutlin <- data_nutlin()

# Take a look at this tibble
glimpse(nutlin)

# Clean data for next steps
nutlin_clean <-
  nutlin %>%
  # Add DepMap IDs
  add_ids(cell = "Cell line") %>%
  # Remove cell lines that weren't matched to an ID
  filter(!is.na(depmap_id)) %>%
  # Remove cell lines without AUC values
  filter(!is.na(AUC))

glimpse(nutlin_clean)

# Correlate gene expression with nutlin-3 AUC
exp_result <- 
  nutlin_clean %>%
  cor_expression(
    response = "AUC",
    ids = "depmap_id"
    )

glimpse(exp_result)

# Merge expression correlations with input nutlin-3 data
exp_merged <- 
  nutlin_clean %>%
  inner_join(data_expression(), by = "depmap_id") %>%
  left_join(exp_result, by = "gene")

glimpse(exp_merged)

# Correlate gene expression with nutlin-3 AUC
mut_result <- 
  nutlin_clean %>%
  cor_mutations(
    response = "AUC",
    ids = "depmap_id"
  )

glimpse(mut_result)

# Merge expression correlations with input nutlin-3 data
mut_merged <- 
  nutlin_clean %>%
  inner_join(data_mutations(), by = "depmap_id") %>%
  left_join(mut_result, by = "gene")

glimpse(mut_merged)

# Plot top gene expression biomarkers
exp_merged %>%
  # Filter for 12 genes with strongest negative correlation
  filter(dense_rank(rho) <= 12) %>%
  # Convert gene to a factor so that faceted plot is sorted by rho
  mutate(gene = fct_reorder(gene, rho)) %>%
  # Create plot
  ggplot(aes(x = rna_expression, y = AUC)) +
  geom_point(alpha = 0.2) +
  geom_smooth(method = "lm", se = FALSE) +
  xlab("RNA expression (log2[TPM + 1])") +
  ylab("Area under curve") +
  ggtitle("Top sensitizing genes") +
  facet_wrap(~gene, scales = "free") +
  theme_bw()

# Plot top mutation biomarkers
mut_merged %>%
  # Plot only significant mutations
  filter(significant) %>%
  # Convert gene to a factor so that faceted plot is sorted by p.value
  mutate(gene = fct_reorder(gene, p.value)) %>%
  # Change naming of mutant to be more explicit
  mutate(mutant = ifelse(mutant, "Mutant", "Wild-type")) %>%
  mutate(mutant = factor(mutant, levels = c("Wild-type", "Mutant"))) %>%
  # Create plot
  ggplot(aes(x = mutant, y = AUC)) +
  geom_boxplot(outlier.shape = NA) +
  geom_jitter(aes(color = mutant), width = 0.2, alpha = 0.4) +
  facet_wrap(~gene) +
  scale_color_viridis_d(option = "C", end = 0.8) +
  theme_bw() +
  theme(legend.position = "none") +
  xlab("Genotype") +
  ylab("Area under curve") +
  ggtitle("Significant mutation biomarkers")