In rkawalerski/pdacR: Obtains and analyzes PDAC transcriptomics data

Setup

load required libraries

library(pdacR)
library(ggplot2)
library(ggpubr)
library(ggrepel)
library(pROC)
library(pdacmolgrad)
library(plyr)

Preprocessing of genesets

dataset <- Moffitt_GEO_array

## Calculate signature scores and SSC
# =====================================================================
# perform single sample classifier for guidance

tumor.classifier <- Moffitt_classifier_2019

dataset$sampInfo$SST_subtypes<- as.numeric(create.classif(dat=dataset$ex,
                                              fit=tumor.classifier$fit,
                                              classifier=tumor.classifier)$predprob)

gene_lists$ADEX_unique <- gene_lists$ICGC.SAM$symbols[which(gene_lists$ICGC.SAM$type %in% "ADEX")]
gene_lists$Immunogenic_unique <- gene_lists$ICGC.SAM$symbols[which(gene_lists$ICGC.SAM$type %in% "Immunogenic")]
gene_lists$Progenitor_unique <- gene_lists$ICGC.SAM$symbols[which(gene_lists$ICGC.SAM$type %in% "Pancreatic progenitor")]
gene_lists$Squamous_unique <- gene_lists$ICGC.SAM$symbols[which(gene_lists$ICGC.SAM$type %in% "Squamous")]



# =====================================================================
# Calculate expression scores
for(i in names(gene_lists)){
  this_gene_list <- gene_lists[[i]]
  if(class(this_gene_list) %in% "data.frame"){
    this_gene_list <- this_gene_list[,1]
  }
  tmp <- which(dataset$featInfo$SYMBOL %in% this_gene_list)
  dataset$sampInfo[i] <- colMeans((dataset$ex[tmp,]),na.rm = TRUE)
}




# =====================================================================
# Make utility tracks
dataset$sampInfo$combined <- paste(dataset$sampInfo$location,
                                   dataset$sampInfo$specimen_type)

dataset$sampInfo$simpleLocation <- as.character(dataset$sampInfo$location)
dataset$sampInfo$simpleLocation[dataset$sampInfo$specimen_type %in% "cell line"] <- "plastic"
dataset$sampInfo$simpleLocation[dataset$sampInfo$location %in% 
                                  c("AbWall","Duo","Colon","Peritoneal",
                                    "Diaphragm","Fat","PelvicWall","Vessel")] <- "Peritoneal/GI"
dataset$sampInfo$simpleLocation[dataset$sampInfo$location %in% 
                                  c("Spleen","LymphNode")] <- "Spleen/LN"
# print(unique(dataset$sampInfo$simpleLocation))

# =====================================================================
# Append purIST and molgrad predictions
temp <- projectMolGrad(dataset$ex,
                       geneSymbols = dataset$featInfo$SYMBOL)
names(temp) <- paste0('molgrad_', names(temp))
temp$geo_accession <- rownames(temp)
# print(head(temp))

dataset$sampInfo <- join(dataset$sampInfo,
                         temp,
                         by = 'geo_accession')
rownames(dataset$ex) <- dataset$featInfo$SYMBOL
dataset$sampInfo$purIST <- as.numeric(create.classif(dataset$ex,
                                                     Moffitt_classifier_2019,
                                                     fit = Moffitt_classifier_2019$fit)$predprob)
dataset$sampInfo$molgrad_scaled <- GGally::rescale01(dataset$sampInfo$molgrad_PDX)

# =====================================================================
# Identify sample subsets
pt_with_primary <-  unique(dataset$sampInfo$ptIndex[
  dataset$sampInfo$specimen_type %in% "Primary"])
pt_with_met <-  unique(dataset$sampInfo$ptIndex[
  dataset$sampInfo$specimen_type %in% "Metastasis"])
pt_with_both <- intersect(pt_with_primary,pt_with_met)

Match samples from same source

data <- droplevels(subset(dataset$sampInfo,ptIndex %in% pt_with_both))
# print(names(data))

# Manually order patient indices for nice plotting
data$ptIndex_f <- factor(data$ptIndex, levels = c("171", "176", "184", "181", "172", "183", "8", "13", "12", "14"))

Results

Exocrine signature by matched samples

ggplot(data,aes(y = Collisson.Exocrine,
                x = combined,
                shape = specimen_type,
                fill = simpleLocation)) +
  geom_point(size=3,alpha=0.8, aes(reorder(combined, Collisson.Exocrine)))  +
  facet_grid(. ~ ptIndex_f) +
  theme_pubr() +
  scale_shape_manual(values = c(21,22,24)) +
  scale_fill_manual(values = c("Lung" = "#B79F00",
                               "Liver" = "#F8766D",
                               "Pancreas" = "#619CFF",
                               "Peritoneal/GI" = "#00BFC4",
                               "Spleen/LN" = "#F564E3")) +
  theme(axis.text.x = element_blank(),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank()) +
  labs(title = "Expression in matched samples from Moffitt Array data",
       x = "Patient ID")

QM signature by matched samples

ggplot(data,aes(y = Collisson.QM,
                x = combined,
                shape = specimen_type,
                fill = simpleLocation)) +
  geom_point(size=3,alpha=0.8)  +
  facet_grid(. ~ ptIndex_f) +
  theme_pubr() +
  scale_shape_manual(values = c(21,22,24)) +
  scale_fill_manual(values = c("Lung" = "#B79F00",
                               "Liver" = "#F8766D",
                               "Pancreas" = "#619CFF",
                               "Peritoneal/GI" = "#00BFC4",
                               "Spleen/LN" = "#F564E3")) +
  theme(axis.text.x = element_blank(),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank()) +
  labs(title = "Expression in matched samples from Moffitt Array data",
       x = "Patient ID")

Collisson Classical signature by matched samples

ggplot(data,aes(y = Collisson.Classical,
                x = combined,
                shape = specimen_type,
                fill = simpleLocation)) +
  geom_point(size=3,alpha=0.8)  +
  facet_grid(. ~ ptIndex_f) +
  theme_pubr() +
  scale_shape_manual(values = c(21,22,24)) +
  scale_fill_manual(values = c("Lung" = "#B79F00",
                               "Liver" = "#F8766D",
                               "Pancreas" = "#619CFF",
                               "Peritoneal/GI" = "#00BFC4",
                               "Spleen/LN" = "#F564E3")) +
  theme(axis.text.x = element_blank(),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank()) +
  labs(title = "Expression in matched samples from Moffitt Array data",
       x = "Patient ID")

ADEX signature by matched samples

ggplot(data,aes(y = ICGC.ADEX.Up,
                x = combined,
                shape = specimen_type,
                fill = simpleLocation)) +
  geom_point(size=3,alpha=0.8)  +
  facet_grid(. ~ ptIndex_f) +
  theme_pubr() +
  scale_shape_manual(values = c(21,22,24)) +
  scale_fill_manual(values = c("Lung" = "#B79F00",
                               "Liver" = "#F8766D",
                               "Pancreas" = "#619CFF",
                               "Peritoneal/GI" = "#00BFC4",
                               "Spleen/LN" = "#F564E3")) +
  theme(axis.text.x = element_blank(),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank()) +
  labs(title = "Expression in matched samples from Moffitt Array data",
       x = "Patient ID")

Immunogenic signature by matched samples

ggplot(data,aes(y = ICGC.Immunogenic.Up,
                x = combined,
                shape = specimen_type,
                fill = simpleLocation)) +
  geom_point(size=3,alpha=0.8)  +
  facet_grid(. ~ ptIndex_f) +
  theme_pubr() +
  scale_shape_manual(values = c(21,22,24)) +
  scale_fill_manual(values = c("Lung" = "#B79F00",
                               "Liver" = "#F8766D",
                               "Pancreas" = "#619CFF",
                               "Peritoneal/GI" = "#00BFC4",
                               "Spleen/LN" = "#F564E3")) +
  theme(axis.text.x = element_blank(),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank()) +
  labs(title = "Expression in matched samples from Moffitt Array data",
       x = "Patient ID")

Basal-like signature by matched samples

ggplot(data,aes(y = Moffitt.Basal.25,
                x = combined,
                shape = specimen_type,
                fill = simpleLocation)) +
  geom_point(size=3,alpha=0.8)  +
  facet_grid(. ~ ptIndex_f) +
  theme_pubr() +
  scale_shape_manual(values = c(21,22,24)) +
  scale_fill_manual(values = c("Lung" = "#B79F00",
                               "Liver" = "#F8766D",
                               "Pancreas" = "#619CFF",
                               "Peritoneal/GI" = "#00BFC4",
                               "Spleen/LN" = "#F564E3")) +
  theme(axis.text.x = element_blank(),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank()) +
  labs(title = "Expression in matched samples from Moffitt Array data",
       x = "Patient ID")

Classical signature by matched samples

ggplot(data,aes(y = Moffitt.Classical.25,
                x = combined,
                shape = specimen_type,
                fill = simpleLocation)) +
  geom_point(size=3,alpha=0.8)  +
  facet_grid(. ~ ptIndex_f) +
  theme_pubr() +
  scale_shape_manual(values = c(21,22,24)) +
  scale_fill_manual(values = c("Lung" = "#B79F00",
                               "Liver" = "#F8766D",
                               "Pancreas" = "#619CFF",
                               "Peritoneal/GI" = "#00BFC4",
                               "Spleen/LN" = "#F564E3")) +
  theme(axis.text.x = element_blank(),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank()) +
  labs(title = "Expression in matched samples from Moffitt Array data",
       x = "Patient ID")

Simplified patient matched sample plots

good_samples <- which(data$ptIndex_f %in% c("8", "12", "13", "14"))
metastasis <- which((data$specimen_type %in% "Metastasis") & (!data$simpleLocation %in% "Pancreas"))
adjacent_normal <- which((data$specimen_type %in% "Normal") & (!data$simpleLocation %in% "Pancreas"))
normal <- which((data$specimen_type %in% "Normal") & (data$simpleLocation %in% "Pancreas"))
primary <- which((data$specimen_type %in% "Primary") & (data$simpleLocation %in% "Pancreas"))

data$simplefactor <- character(nrow(data))
data$simplefactor[metastasis] <- "Metastasis"
data$simplefactor[adjacent_normal] <- "Adj.normal"
data$simplefactor[normal] <- "Normal"
data$simplefactor[primary] <- "Primary"
data$simplefactor <- as.factor(data$simplefactor)

levels(data$simplefactor) <- c("Metastasis", "Adj.normal", "Normal", "Primary")

for(genes in noquote(c(names(gene_lists),
                       "molgrad_PDX",
                        "molgrad_Puleo",
                        "molgrad_ICGCarray",
                        "molgrad_ICGCrnaseq",
                        "purIST",
                        "molgrad_scaled"))){

  d <- data.frame(score = scale(data[good_samples, genes],
                                center = TRUE,
                                scale = TRUE),
                  location = data$simplefactor[good_samples],
                  type = data$specimen_type[good_samples])

  p <- ggplot(d, aes(y = score,
                     x = location)) +
    geom_dotplot(aes(fill = location),
                 binaxis='y', 
                 stackdir='center', 
                 binwidth = 0.115,
                 dotsize=1.25) +
    stat_compare_means(ref.group = "Primary") +
    ylim(-2,4) +

    labs(title = paste(genes, " score")) +
    theme_pubr()


  print(p)

}

Simplified patient matched sample plots, all samples

metastasis <- which((dataset$sampInfo$specimen_type %in% "Metastasis") & (!dataset$sampInfo$simpleLocation %in% "Pancreas"))
adjacent_normal <- which((dataset$sampInfo$specimen_type %in% "Normal") & (!dataset$sampInfo$simpleLocation %in% "Pancreas"))
normal <- which((dataset$sampInfo$specimen_type %in% "Normal") & (dataset$sampInfo$simpleLocation %in% "Pancreas"))
primary <- which((dataset$sampInfo$specimen_type %in% "Primary") & (dataset$sampInfo$simpleLocation %in% "Pancreas"))


dataset$sampInfo$simplefactor <- character(nrow(dataset$sampInfo))
dataset$sampInfo$simplefactor[metastasis] <- "Metastasis"
dataset$sampInfo$simplefactor[adjacent_normal] <- "Adj.normal"
dataset$sampInfo$simplefactor[normal] <- "Normal"
dataset$sampInfo$simplefactor[primary] <- "Primary"
dataset$sampInfo$simplefactor[which(dataset$sampInfo$specimen_type %in% "cell line")] <- "Cell line"
dataset$sampInfo$simplefactor <- as.factor(dataset$sampInfo$simplefactor)


dataset$sampInfo$simplefactor <- factor(dataset$sampInfo$simplefactor,
                                        levels = c("Cell line", "Metastasis", "Adj.normal", "Normal", "Primary"))

dataset$sampInfo$normal_lung_adex <- character(length = nrow(dataset$sampInfo))
dataset$sampInfo$normal_lung_adex[which(dataset$sampInfo$UNC_MADB_ID %in% "48607")] <- "yes"
dataset$sampInfo$normal_lung_adex <- as.factor(dataset$sampInfo$normal_lung_adex)

noscale <- c("molgrad_PDX",
                        "molgrad_Puleo",
                        "molgrad_ICGCarray",
                        "molgrad_ICGCrnaseq",
                        "purIST",
                        "molgrad_scaled")


for(genes in noquote(c(names(gene_lists),
                       "molgrad_PDX",
                        "molgrad_Puleo",
                        "molgrad_ICGCarray",
                        "molgrad_ICGCrnaseq",
                        "purIST",
                        "molgrad_scaled"))){
  if(genes %in% noscale){
    d <- data.frame(score = dataset$sampInfo[, genes],
                  location = dataset$sampInfo$simpleLocation,
                  type = dataset$sampInfo$simplefactor,
                  outlier = dataset$sampInfo$normal_lung_adex)

  p <- ggplot(d, aes(y = score,
                     x = type)) +
    geom_dotplot(aes(fill = location),
                 binaxis='y', 
                 stackdir='center', 
                 binwidth = 0.05,
                 dotsize=0.45,
                 method = 'histodot',
                 position = position_dodge(1)) +
    geom_text_repel(data = subset(d, outlier %in% "yes"),
                    aes(label = outlier)) +
    stat_compare_means(ref.group = "Primary") +
    labs(title = paste(genes, " score")) +
    theme_pubr()

  print(p)


  }else{
    d <- data.frame(score = scale(dataset$sampInfo[, genes],
                                center = TRUE,
                                scale = TRUE),
                  location = dataset$sampInfo$simpleLocation,
                  type = dataset$sampInfo$simplefactor,
                  outlier = dataset$sampInfo$normal_lung_adex)

  p <- ggplot(d, aes(y = score,
                     x = type)) +
    geom_dotplot(aes(fill = location),
                 binaxis='y', 
                 stackdir='center', 
                 binwidth = 0.05,
                 dotsize=1.5,
                 method = 'histodot',
                 position = position_dodge(1)) +
    geom_text_repel(data = subset(d, outlier %in% "yes"),
                    aes(label = outlier)) +
    stat_compare_means(ref.group = "Primary") +
    ylim(-3,4) +
    labs(title = paste(genes, " score")) +
    theme_pubr()


  print(p)
  }
}

rkawalerski/pdacR documentation built on Jan. 25, 2025, 10:50 a.m.