Add response to JACI review and update figures

3.P259.pDC_response.to.review.Rmd

@@ -0,0 +1,360 @@
+---
+title: 'P259: Response to reviews'
+subtitle: "Dendritic cells (pDC)"
+author: "Kim Dill-McFarland, [email protected]"
+date: "version `r format(Sys.time(), '%B %d, %Y')`"
+output:
+  pdf_document:
+    toc: yes
+    toc_depth: '3'
+  html_document:
+    toc: yes
+    toc_depth: 3
+    toc_float:
+      collapsed: no
+editor_options:
+  chunk_output_type: console
+---
+# Background
+
+Analyses performed in response to reviews for publication in JACI.
+
+# Setup
+Load packages
+
+```{r message=FALSE, warning=FALSE}
+# Data manipulation and figures
+library(tidyverse)
+library(ggpubr)
+library(readxl)
+library(patchwork)
+#Print pretty tables to Rmd
+library(knitr)
+library(kableExtra)
+```
+
+Set seed
+
+```{r}
+set.seed(589)
+```
+
+Custom functions
+
+```{r}
+#To extract and format p-value from lmFit
+source("https://raw.githubusercontent.com/kdillmcfarland/R_bioinformatic_scripts/master/limma.extract.pval.R")
+```
+
+# Low-quality libraries
+
+Determine if low-quality libraries removed from analysis are associated with low input RNA. All filtered libraries with low raw sequences and high CV coverage also had low input RNA.
+
+```{r echo=FALSE, message=FALSE, fig.width=8.5}
+#### Data ####
+attach("data_clean/P259_pDC_clean.RData")
+meta <- read_csv("data_clean/P259_pDC_metadata.csv")
+anno <- read_excel("data_raw/P259-2 Final Annotation.xlsx")
+
+dat <- meta %>% 
+  select(libID, total_sequences, median_cv_coverage) %>% 
+  inner_join(select(anno, "library sampleId", "RNA conc (ng/ul)"), 
+             by=c("libID"="library sampleId")) %>% 
+  mutate(`Quality filter` = ifelse(libID %in% dat.pDC.voom$targets$libID, "Pass", "Fail"))
+
+#### Plot ####
+plot1 <- dat %>% 
+  ggplot(aes(y=`RNA conc (ng/ul)`, x=total_sequences)) +
+  geom_point(aes(color=`Quality filter`), size=2) +
+  theme_classic() +
+  labs(x = "Raw sequences", y="RNA concentration (ng/ul)")
+
+plot2 <- dat %>% 
+  ggplot(aes(y=`RNA conc (ng/ul)`, x=median_cv_coverage)) +
+  geom_point(aes(color=`Quality filter`), size=2) +
+  theme_classic() +
+  labs(x = "Raw sequences", y="RNA concentration (ng/ul)")
+
+plot1+plot2
+```
+
+# Viral load
+### Data
+
+RV sequences were extracted from RNA-seq libraries and quantified using methods similar to human data. Normalized RV counts were calculated as total RV sequences / total non-human sequences * 1E6
+
+```{r echo=FALSE, message=FALSE}
+dat2 <- read_csv("data_raw/211102_P259-1_P259-2_RhinoViruses_normlized_read_counts.csv") %>% 
+  rename(libID=libid) %>% 
+  #filter libraries in final analysis
+  filter(libID %in% dat.pDC.voom$targets$libID) %>% 
+  #add metadata
+  left_join(select(meta, libID, experiment, donorID, IL5, virus, virus.detail)) %>% 
+  mutate(contrast=paste(IL5,virus, sep="_")) %>% 
+  mutate(contrast=factor(contrast, 
+                         levels = c("none_none","none_HRV",
+                                 "AntiIL5_none","AntiIL5_HRV",
+                                 "EOS.supp_none","EOS.supp_HRV")))
+```
+
+```{r echo=FALSE, message=FALSE}
+# Format for limma
+count1 <- dat2 %>% 
+  filter(experiment == "P259_1") %>% 
+  select(libID, RhinoVirusA_normCount) %>% 
+  pivot_longer(-libID) %>% 
+  arrange(libID) %>% 
+  pivot_wider(names_from = libID) %>% 
+  column_to_rownames("name")
+
+meta1 <- dat2 %>% 
+  filter(experiment == "P259_1") %>% 
+  arrange(libID) %>% 
+  droplevels()
+
+count2 <- dat2 %>% 
+  filter(experiment == "P259_2") %>% 
+  select(libID, RhinoVirusA_normCount) %>% 
+  pivot_longer(-libID) %>% 
+  arrange(libID) %>% 
+  pivot_wider(names_from = libID) %>% 
+  column_to_rownames("name")
+
+meta2 <- dat2 %>% 
+  filter(experiment == "P259_2") %>% 
+  arrange(libID) %>% 
+  droplevels()
+```
+
+### Run limma contrasts model
+
+In both experiments, viral loads were significantly higher in virus-infected samples and there were no differences between EOS supernatant or AntiIL5 treatment groups.
+
+```{r echo=FALSE}
+# Define model
+model_1.contrast<- model.matrix(~ 0 + contrast, data=meta1)
+    colnames(model_1.contrast) <- c(
+                                 "none_none","none_HRV",
+                                 "EOS.supp_none","EOS.supp_HRV")
+#Block by donor  
+consensus.corr1 <- duplicateCorrelation(
+  count1,
+  model_1.contrast,
+  block=meta1$donorID)$consensus.correlation
+
+#Fit model
+fitQW_1.contrast <- lmFit(count1, 
+                      model_1.contrast,
+                      block=meta1$donorID,
+                      correlation=consensus.corr1)
+
+#Get contrasts
+contrast.matrix1 <- makeContrasts(
+  none_HRV-none_none, 
+  EOS.supp_HRV-EOS.supp_none,
+  EOS.supp_none-none_none,
+  EOS.supp_HRV-none_HRV,
+  levels=model_1.contrast)
+
+efitQW_1.contrast <- eBayes(contrasts.fit(fitQW_1.contrast,
+                                  contrast.matrix1))
+
+extract.pval(model=model_1.contrast,
+             voom.dat=count1, 
+             eFit=efitQW_1.contrast, 
+             name="pval_1.contrast",
+             summary=FALSE,
+             contrasts=TRUE,
+             contrast.mat=contrast.matrix1)
+```
+
+```{r echo=FALSE}
+# Define model
+model_2.contrast<- model.matrix(~ 0 + contrast, data=meta2)
+    colnames(model_2.contrast) <- c(
+                                 "none_none","none_HRV",
+                                 "AntiIL5_none","AntiIL5_HRV")
+#Block by donor  
+consensus.corr2 <- duplicateCorrelation(
+  count2,
+  model_2.contrast,
+  block=meta2$donorID)$consensus.correlation
+
+#Fit model
+fitQW_2.contrast <- lmFit(count2, 
+                      model_2.contrast,
+                      block=meta2$donorID,
+                      correlation=consensus.corr2)
+
+#Get contrasts
+contrast.matrix2 <- makeContrasts(
+  none_HRV-none_none, 
+  AntiIL5_HRV-AntiIL5_none,
+  AntiIL5_none-none_none,
+  AntiIL5_HRV-none_HRV,
+  levels=model_2.contrast)
+
+efitQW_2.contrast <- eBayes(contrasts.fit(fitQW_2.contrast,
+                                  contrast.matrix2))
+
+extract.pval(model=model_2.contrast,
+             voom.dat=count2, 
+             eFit=efitQW_2.contrast, 
+             name="pval_2.contrast",
+             summary=FALSE,
+             contrasts=TRUE,
+             contrast.mat=contrast.matrix2)
+```
+
+```{r echo=FALSE}
+pval_1.contrast %>% 
+  select(group, logFC, adj.P.Val) %>% 
+  kable(align=c("l","c","c"), digits=4, caption="P259_1",
+        col.names = c("Variable","Fold change", "FDR")) %>% 
+  kable_styling(latex_options="HOLD_position", full_width = FALSE)
+
+pval_2.contrast %>% 
+  select(group, logFC, adj.P.Val) %>% 
+  kable(align=c("l","c","c"), digits=4, caption="P259_2", 
+        col.names = c("Variable","Fold change", "FDR")) %>% 
+  kable_styling(latex_options="HOLD_position", full_width = FALSE)
+```
+
+```{r echo=FALSE}
+#save
+write_csv(pval_1.contrast, file="review_response/RV.EOS.model.csv")
+write_csv(pval_2.contrast, file="review_response/RV.antiIL5.model.csv")
+```
+
+### Plot
+
+```{r echo=FALSE, fig.width=8.5, message=FALSE, warning=FALSE}
+#### Format pval data ####
+#Pvals for eos experiment
+GOI.p1 <- pval_1.contrast %>% 
+  mutate(dataset="P259.1")
+
+#Pvals for aniti-IL5 experiment
+GOI.p <- pval_2.contrast%>% 
+  mutate(dataset="P259.2") %>% 
+  #Combine with eos data
+  bind_rows(GOI.p1) %>% 
+  #Make symbols for plots
+  mutate(symbol = ifelse(P.Value <= 0.001,"***",
+                         ifelse(P.Value <= 0.01, "**",
+                                ifelse(P.Value <= 0.05, "*", NA)))) %>% 
+  filter(!is.na(symbol)) %>% 
+  #Match group labels to those in plots
+  mutate(group.long = paste(dataset, group, sep="_"),
+         group1 = recode_factor(factor(group.long),
+                        "P259.2_none_HRV - none_none"='"-\n-"',
+                        "P259.2_AntiIL5_HRV - AntiIL5_none"='"-\n+"',
+                        "P259.2_AntiIL5_none - none_none"='"-\n-"',
+                        "P259.2_AntiIL5_HRV - none_HRV"='"+\n-"',
+                        "P259.1_none_HRV - none_none"='"RV            -\nEOS sup  -"',
+                        "P259.1_EOS.supp_HRV - EOS.supp_none"='"-\n+"',
+                        "P259.1_EOS.supp_none - none_none"='"RV            -\nEOS sup  -"',
+                        "P259.1_EOS.supp_HRV - none_HRV"='"+\n-"'),
+         group2 = recode_factor(factor(group.long),
+                          "P259.2_none_HRV - none_none"='"+\n-"',
+                          "P259.2_AntiIL5_HRV - AntiIL5_none"='"+  RV\n+  Anti-IL-5/5R"*alpha',
+                          "P259.2_AntiIL5_none - none_none"='"-\n+"',
+                          "P259.2_AntiIL5_HRV - none_HRV"='"+  RV\n+  Anti-IL-5/5R"*alpha',
+                          "P259.1_none_HRV - none_none"='"+\n-"',
+                          "P259.1_EOS.supp_HRV - EOS.supp_none"='"+\n+"',
+                          "P259.1_EOS.supp_none - none_none"='"-\n+"',
+                          "P259.1_EOS.supp_HRV - none_HRV"='"+\n+"'),
+         facet.lab = recode_factor(factor(dataset),
+                            "P259.1" = '"EOS sup"',
+                            "P259.2" = '"Anti-IL-5/5R"*alpha')) %>% 
+  dplyr::select(facet.lab, group1, group2, symbol)
+
+#Add y location for pval based on max expression in plot
+GOI.pe <- dat2 %>% 
+  #Max expression per gene and experiment
+  group_by(experiment) %>% 
+  summarise(max.e = max(RhinoVirusA_normCount, na.rm=TRUE)) %>% 
+  ungroup() %>% 
+  mutate(facet.lab = experiment,
+         facet.lab = recode_factor(factor(facet.lab),
+                            "P259_1" = '"EOS sup"',
+                            "P259_2" = '"Anti-IL-5/5R"*alpha')) %>% 
+  #Add to pval data
+  right_join(GOI.p) %>% 
+  arrange(experiment, group1, group2)
+
+#first entry per gene
+#Set y position to 1
+first <- GOI.pe %>% 
+  group_by(experiment) %>% 
+  slice(1) %>% 
+  mutate(y.position1 = 1)
+
+#Add first position data back and fill in remaining
+GOI.pey <- GOI.pe %>% 
+  full_join(first) %>% 
+  
+  #Fill in positions 2 - N
+  group_by(experiment, facet.lab) %>% 
+  mutate(y.position2 = lag(y.position1)+1,
+         y.position3 = lag(y.position2)+1,
+         y.position4 = lag(y.position3)+1) %>% 
+  #Collapse positions into 1 column
+  mutate(y.position = ifelse(!is.na(y.position1),y.position1,
+                             ifelse(!is.na(y.position2),y.position2,
+                                    ifelse(!is.na(y.position3),y.position3,
+                                           ifelse(!is.na(y.position4),y.position4,NA))))) %>% 
+  #Scale to max expression
+  mutate(y.position = max.e+y.position*5E3) %>% 
+  ungroup()
+#plot
+plot3 <- dat2 %>% 
+  mutate(x.lab=paste(experiment, virus.detail, IL5, sep="_"),
+         x.lab=gsub("oldH|newH", "", x.lab),
+         x.lab=recode_factor(factor(x.lab),
+                             "P259_1_none_none"='"RV            -\nEOS sup  -"',
+                             "P259_2_none_none"='"-\n-"',
+                             "P259_1_none_EOS.supp"='"-\n+"',
+                             "P259_2_none_AntiIL5"='"-\n+"',
+                             "P259_1_RV_none"='"+\n-"',
+                             "P259_2_RV_none"='"+\n-"',
+                             "P259_1_RV_EOS.supp"='"+\n+"',
+                             "P259_2_RV_AntiIL5"='"+  RV\n+  Anti-IL-5/5R"*alpha')) %>% 
+  mutate(facet.lab = experiment,
+         facet.lab = recode_factor(factor(facet.lab),
+                            "P259_1" = '"EOS sup"',
+                            "P259_2" = '"Anti-IL-5/5R"*alpha')) %>% 
+  arrange(facet.lab) %>% 
+  ggplot(aes(x=x.lab, y=RhinoVirusA_normCount, color=donorID)) +
+  geom_jitter(width=0.1, height=0) +
+  stat_summary(fun.data=mean_sdl, 
+               fun.args = list(mult=1), 
+               geom="errorbar", color="black", width=0.25) +
+  stat_summary(fun=mean, geom="errorbar", 
+               aes(ymax=..y.., ymin=..y..),
+               color="black", width=0.5) +
+  facet_grid(~facet.lab, scales="free", 
+             labeller = labeller(facet.lab=label_parsed)) +
+  #Add pval
+  stat_pvalue_manual(data=GOI.pey,
+                     label="symbol", xmin="group1", xmax="group2") +
+  # #Beautify
+  theme_bw() +
+  labs(x="", y="Normalized RV expression") +
+  theme(legend.position = "none",
+        panel.grid.major.y = element_blank(),
+        panel.grid.minor.y = element_blank(),
+        strip.background =element_rect(fill="white"),
+        axis.text.x=element_text(hjust=0.95,vjust=-0.5),
+        plot.margin = margin(0.2,0.2,0.2,0.2,"cm")) +
+  scale_x_discrete(labels = ggplot2:::parse_safe)
+plot3
+```
+
+# R session
+
+```{r}
+sessionInfo()
+```
+
+***