Figure 2.Rmd

---
title: "Differential Expression Analysis Day 0 Samples"
author: "Megan Hockman"
date: "Last compiled on `r format(Sys.time(), '%d %B, %Y')`"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

```{r}
library(extrafont)
library(remotes)
remotes::install_version("Rttf2pt1", version = "1.3.8")
extrafont::font_import()
```

```{r}
library(biomaRt)
library(cowplot)
library(clusterProfiler)
# library(DESeq2)
library(edgeR)
library(enrichplot)
library(gameofthrones)
library(ggpubr)
library(GOSemSim)
library(ggplot2)
library(ggridges)
# library(gprofiler2)
# library(openxlsx)
# library(pathview)
library(reshape2)
# library(see)
library(splines)
# library(SuperExactTest)
library(UpSetR)
library(png)
library(ggpubr)
library(dplyr)
library(stringr) 
library(gridExtra)
library(egg)
library(patchwork)
library(extrafont)
```

###Load the Data and log2 seroconversion
```{r}
load("../data.RData")

metadata$log2sero <- log2(metadata$average_seroconversion)

metadata$log2baseline <- log2(metadata$Composite_Baseline)

```

## Check Covariates
This code makes histograms of frequencies within covariate terms, to determine if any term is underrepresented in the metadata.

```{r}
covariates <- c("Age", "Gender", "BMI_Value", "Race_Ethnicity", "log2baseline", "log2sero", "Month.vaccinated", "Prevacc.naive")

column <- c()
for (i in 1:length(covariates)){

  column[[i]] <- which(colnames(metadata)==covariates[i])  
  
}

as.character(column)
histolist <- list()

for (i in 1:length(column)) {
  histolist[[i]] <- ggplot(data = metadata, aes(x = !!metadata[,column[[i]]])) +
    geom_bar(stat = "count") +
    geom_text(stat='count', aes(label=..count..), vjust=-1) +
    labs(x = covariates[[i]],
         y = "counts")
}

histolist

```

##Set covariates of interest
These include covariates that are well distributed across the subjects:
1) BMI
2) Gender
3) Age
4) Average seroconversion
5) Baseline HAI
```{r}
covariates <- c("BMI_Value", "Gender", "Age", "log2sero", "log2baseline")

covariates_titles <- c("BMI", "Sex", "Age", "Seroconversion", "Baseline HAI")

#metadata$Seroconversion_bivar <- ifelse(metadata$Composite_SeroConversion > 8, "high", "low")


fig2 <- list(a = c(),
             b = c(),
             c = c(),
             d = c(),
             e = c(),
             f = c())
```


Calculate Differential Expression based on selected covariates
```{r}
FDR.cutoff <- 0.05
degenes.edgeR <- list()
toptags <- list()

for(j in 1:length(covariates)){
  group <- metadata[,which(colnames(metadata) == covariates[j])]
  
  if(length(unique(group)) > 10){
    spline <- ns(group, df = 1)
    design <- model.matrix(~spline)
  }else{
    design <- model.matrix(formula(paste("~",covariates[j])),data = metadata)
  }
  
  y <- DGEList(counts = counts.subset,
               group = group)
  
  keep <- filterByExpr(y)
  
  y <- y[keep, ,keep.lib.sizes = F]
  y <- calcNormFactors(y, method = "TMM")
  y <- estimateDisp(y, design, robust = TRUE)
  fit <- glmQLFit(y, design, robust = TRUE)
  
  qlf <- glmQLFTest(fit, coef = 2:ncol(design))
  degenes <- topTags(qlf, n = Inf)
  toptags[[j]] <- degenes
  degenes.edgeR[[j]] <- rownames(degenes$table)[degenes@.Data[[1]]$FDR < FDR.cutoff]
  
}

names(degenes.edgeR) <- covariates_titles

```

Plot UpSet plot of Differentially expressed genes

```{r}
fig2$a<- upset(fromList(degenes.edgeR), 
                order.by = "freq", 
                nsets = length(covariates),
                mainbar.y.max = 3000, 
                text.scale = 5, 
                mainbar.y.label = "Intersection Size", 
                sets.x.label = "Differentially\nExpressed Genes", 
                point.size = 6,
                line.size = 1, 
                nintersects = 12,
               set_size.show = TRUE)

fig2$a

pdf(paste0("fig2a.pdf"),
    height=15,
    width=30)
fig2$a
dev.off()

png(paste0("fig2a.png"),
    height=15,
    width=30, 
    units = "in", 
    res = 300)
fig2$a
dev.off()

```

Export TopTags Tables
```{r}

mart <- useDataset("hsapiens_gene_ensembl", useMart("ensembl"))


for(j in 1:length(toptags)){
  genenames <- as.data.frame(matrix(nrow = nrow(toptags[[j]]),
                                 ncol = 2))
  geneids <- rownames(!!toptags[[j]]$table)
  genenames <- as.data.frame(getBM(filters= "ensembl_gene_id", 
                  attributes = c("ensembl_gene_id", "hgnc_symbol"),
                  values=geneids,
                  mart= mart))
  toptags[[j]]$table$hgnc <- genenames$hgnc_symbol[match(rownames(toptags[[j]]$table), genenames$ensembl_gene_id)]
  print(j)
}


write.csv(toptags[[1]]$table, file = "TopTags_BMI.csv", row.names = TRUE)

write.csv(toptags[[2]]$table, file = "TopTags_Gender.csv", row.names = TRUE)

write.csv(toptags[[3]]$table, file = "TopTags_Age.csv", row.names = TRUE)

write.csv(toptags[[4]]$table, file = "TopTags_log2Seroconversion.csv", row.names = TRUE)

write.csv(toptags[[5]]$table, file = "TopTags_log2BaselineHAI.csv", row.names = TRUE)

```


##Visualize trends of the top x differentially expressed genes within Baseline HAI

1. Get gene information from toptags of interest. Be sure that the index in toptags matches the covariate you're interested in.
2. Filter out degenes below desired false discovery rate threshold.
3. Select the number of genes you want to visualize (default is set to be 50).
4. Get count data that matches differentially expressed genes
5. Make a dataframe containing count data and covariate information (set to baseline HAI value as default)
```{r}
genes <- toptags[[5]]$table
genes <- genes[genes$FDR < 0.05,]
genes <- rownames(genes) #[1:100]



log2tpm.genes <- log2(tpm.subset[genes,])
plot.df <- melt(as.matrix(log2tpm.genes))
plot.df <- plot.df[plot.df[,3] != -Inf,]


x <- dplyr::count(plot.df, vars = Var1)
x <- filter(x, n > 50)
genesig <- as.character(x$vars)
log2tpm.genesig <- log2(tpm.subset[genesig,])
plot.df <- melt(as.matrix(log2tpm.genesig))

plot.df[,4] <- metadata$log2baseline[match(plot.df[,2], metadata$File.Name)]

mart <- useEnsembl(biomart = "ensembl", "hsapiens_gene_ensembl", mirror = "uswest")


plot.list <- list()

for(j in 1:10){
  plot.df.subset <- plot.df[plot.df[,1] == genesig[j],]
  plot.df.subset <- plot.df.subset[plot.df.subset[,3] != -Inf,]
  
  genelist <- getBM(filters= "ensembl_gene_id", 
                  attributes= c("ensembl_gene_id","hgnc_symbol"),
                  values=plot.df.subset$Var1,
                  mart= mart)
  
  plot.df.subset$Var5 <- genelist$hgnc_symbol[match(plot.df.subset$Var1, genelist$ensembl_gene_id)]
  if(is.na(plot.df.subset$Var5)){plot.df.subset$Var5 <- plot.df.subset$Var1[1]}
  
  plot.list[[j]] <- ggplot(data = plot.df.subset, aes(x = !!plot.df.subset[,4],
                                                      y = !!plot.df.subset[,3]
                                                      )) +
    geom_point() +
    geom_smooth(method = "lm",  se = T,color="red") +
    guides(fill = 'none') +
    labs(x = expression("log"[2]*BaselineHAI),
         y = expression("log"[2]*TPM),
         title = plot.df.subset[,5]) +
    theme_bw()+
    theme(plot.title = element_text(size = 10),
          axis.text = element_text(size = 8),
          axis.title = element_text(size = 10),
          aspect.ratio = 1)
  
}

plot.list

png(paste0("figs9.png"),
    height = 5000,
    width = 2000,
    units = 'px',
    res = 300)
plot_grid(plotlist = plot.list, ncol =2, align = 'hv')
dev.off()

pdf(paste0("figs9.pdf"),
    height = 20,
    width = 10)
plot_grid(plotlist = plot.list, ncol = 2, align = 'hv')
dev.off()

```

Visualize trends of the top 50 differentially expressed genes within Seroconversion
```{r}
genes <- toptags[[4]]$table
genes <- genes[genes$FDR < 0.05,]
genes <- rownames(genes)

log2tpm.genes <- log2(tpm.subset[genes,])
plot.df <- melt(as.matrix(log2tpm.genes))
plot.df <- plot.df[plot.df[,3] != -Inf,]

x <- dplyr::count(plot.df, vars = Var1)
x <- filter(x, n > 50)
genesig <- as.character(x$vars)
log2tpm.genesig <- log2(tpm.subset[genesig,])
plot.df <- melt(as.matrix(log2tpm.genesig))

plot.df[,4] <- metadata$log2sero[match(plot.df[,2], metadata$File.Name)]


mart <- useEnsembl(biomart = "ensembl", "hsapiens_gene_ensembl", mirror = "uswest")

plot.list <- list()

for(j in 1:10){
  plot.df.subset <- plot.df[plot.df[,1] == genesig[j],]
  plot.df.subset <- plot.df.subset[plot.df.subset[,3] != -Inf,]
  
  genelist <- getBM(filters= "ensembl_gene_id", 
                  attributes= c("ensembl_gene_id","hgnc_symbol"),
                  values=plot.df.subset$Var1,
                  mart= mart)
  
  plot.df.subset$Var5 <- genelist$hgnc_symbol[match(plot.df.subset$Var1, genelist$ensembl_gene_id)]
  if(is.na(plot.df.subset$Var5)){plot.df.subset$Var5 <- plot.df.subset$Var1[1]}
  
  plot.list[[j]] <- ggplot(data = plot.df.subset, aes(x = (!!plot.df.subset[,4]),
                                                      y = !!plot.df.subset[,3])) +
    geom_point() +
    geom_smooth(method = "lm", se = T,color="red") +
    guides(fill = 'none') +
    labs(x = expression("log"[2]*AvgSeroconversion),
         y = expression("log"[2]*TPM),
         title = plot.df.subset[,5]) +
    theme_bw()+
    theme(plot.title = element_text(size = 10),
          axis.text = element_text(size = 8),
          axis.title = element_text(size = 10),
          aspect.ratio = 1)
  
}

plot.list

png(paste0("figs10.png"),
    height = 5000,
    width = 2000,
    units = 'px',
    res = 300)
plot_grid(plotlist = plot.list, ncol =2, align = 'hv')
dev.off()

pdf(paste0("figs10.pdf"),
    height = 20,
    width = 10)
plot_grid(plotlist = plot.list, ncol = 2, align = 'hv')
dev.off()
```
Select genes to report in figure
###Once you've determined which genes can be shown in a figure (usually ones with noticeable trends), select them and make plots.

###In this code:
###Baseline HAI Genes: 
####Increased expression: IGLV3-25
####Decreased expression: LDOC
###Seroconversion Genes:
####Increased expression: IGLV3-9
####Decreased expression: NRAV

Plot representative differentailly expressed genes with respect to baseline HAI
```{r}
names(degenes.edgeR) <- covariates

genes <- c("ENSG00000211659", "ENSG00000182195")

log2tpm.genes <- log2(tpm.subset[genes,])

plot.df <- melt(as.matrix(log2tpm.genes))
plot.df[,4] <- metadata$log2baseline[match(plot.df[,2], metadata$File.Name)]

mart <- useDataset("hsapiens_gene_ensembl", useMart("ensembl"))

plot.list <- list()

for(j in 1:length(genes)){
  plot.df.subset<- plot.df[plot.df[,1] == genes[j],]
  plot.df.subset <- plot.df.subset[plot.df.subset[,3] != -Inf,]
  
  genelist <- getBM(filters= "ensembl_gene_id", 
                  attributes= c("ensembl_gene_id","hgnc_symbol"),
                  values=plot.df.subset$Var1,
                  mart= mart)
  
  plot.df.subset$Var5 <- genelist$hgnc_symbol[match(plot.df.subset$Var1, genelist$ensembl_gene_id)]
  
   plot.list[[j]] <- ggplot(data = plot.df.subset, aes(x = !!plot.df.subset[,4],
                                                      y = !!plot.df.subset[,3])) +
    geom_point() +
    geom_smooth(method = "lm", se = F,color= "#15968BFF") +
     scale_color_viridis_c() +
    guides(fill = 'none') +
    labs(x = "Baseline HAI",
         y = expression("log"[2]*TPM),
         title = plot.df.subset[,5]) +
    theme_bw()+
    theme(text = element_text(family = "Arial", size = 20),
          aspect.ratio = 1)
  
}

plot.list

fig2$b <- plot.list[[1]]
fig2$c <- plot.list[[2]]

pdf(paste0("fig2b.pdf"),
    height=5,
    width=5)
plot(fig2$b)
dev.off()

pdf(paste0("fig2c.pdf"),
    height=5,
    width=5)
plot(fig2$c)
dev.off()

png(paste0("fig2b.png"),
    height=1200,
    width=1200,
    units = "px",
    res = 300)
plot(fig2$b)
dev.off()

png(paste0("fig2c.png"),
    height=1200,
    width=1200,
    units = "px", 
    res = 300)
plot(fig2$c)
dev.off()
```

Plot representative differentially expressed genes as a function of seroconversion
```{r}
names(degenes.edgeR) <- covariates

genes <- c("ENSG00000248008", "ENSG00000211892")

log2tpm.genes <- log2(tpm.subset[genes,])

plot.df <- melt(as.matrix(log2tpm.genes))
plot.df[,4] <- metadata$log2sero[match(plot.df[,2], metadata$File.Name)]

mart <- useDataset("hsapiens_gene_ensembl", useMart("ensembl"))

plot.list <- list()

for(j in 1:length(genes)){
  plot.df.subset<- plot.df[plot.df[,1] == genes[j],]
  plot.df.subset <- plot.df.subset[plot.df.subset[,3] != -Inf,]
  
  genelist <- getBM(filters= "ensembl_gene_id", 
                  attributes= c("ensembl_gene_id","hgnc_symbol"),
                  values=plot.df.subset$Var1,
                  mart= mart)
  
  plot.df.subset$Var5 <- genelist$hgnc_symbol[match(plot.df.subset$Var1, genelist$ensembl_gene_id)]
  
   plot.list[[j]] <- ggplot(data = plot.df.subset, aes(x = !!plot.df.subset[,4],
                                                      y = !!plot.df.subset[,3])) +
    geom_point() +
    geom_smooth(method = "lm", se = F,color="#15968BFF") + #formula = y ~ splines::bs(x, 3),
    guides(fill = 'none') +
    labs(x = expression("Seroconversion"),
         y = expression("log"[2]*TPM),
         title = plot.df.subset[,5]) +
    theme_bw()+
    theme(text = element_text(size = 20, family = "Arial"),
          aspect.ratio = 1)
  
}

plot.list

fig2$d <- plot.list[[1]]
fig2$e <- plot.list[[2]]

pdf(paste0("fig2d.pdf"),
    height=5,
    width=5)
plot(fig2$d)
dev.off()

pdf(paste0("fig2e.pdf"),
    height=5,
    width=5)
plot(fig2$e)
dev.off()

png(paste0("fig2d.png"),
    height=1200,
    width=1200,
    units = "px",
    res = 300)
plot(fig2$d)
dev.off()

png(paste0("fig2e.png"),
    height=1200,
    width=1200,
    units = "px",
    res = 300)
plot(fig2$e)
dev.off()
```

Perform GSEA analysis of genes enriched with respect to baseline HAI

## Download human annotations for GSEA

```{r,warning=FALSE,message = FALSE}
organism = "org.Hs.eg.db"
#BiocManager::install(organism, character.only = TRUE)
library(organism, character.only = TRUE)
```


#This code generates a list of differentially expressed genes, ranked by FC, with respect to baseline HAI
```{r}
target.genes <- toptags[[5]]$table
target.genes <- target.genes[target.genes$FDR < 0.05,]



gene.list <- target.genes$logFC
names(gene.list) <- rownames(target.genes)
gene.list <-  sort(gene.list, decreasing = TRUE)
```


Run GSEA
```{r}
ontlist <- c("BP", "CC", "MF")
simplist <- list()
resultlist <- list()
tib.list <- list()

for (i in 1:length(ontlist)){
  hsGO <- godata('org.Hs.eg.db', ont = ontlist[i])
  set.seed(86)
  gse <- gseGO(geneList=gene.list, 
             ont = ontlist[i], 
             keyType = "ENSEMBL", 
             minGSSize = 3, 
             maxGSSize = 800, 
             pvalueCutoff = 0.05,
             verbose = TRUE, 
             OrgDb = get(organism), 
             pAdjustMethod = "none",
             seed=T)

  simp <- clusterProfiler::simplify(gse, 
                                    semData = hsGO,
                                    cutoff = 0.7)
  simplist[[i]] <- simp
  tib.list[[i]] <- as_tibble(simp@result)
}


gsea.diff.expression_HAI <- as_tibble(rbind(tib.list[[1]], tib.list[[2]], tib.list[[3]]))

gsea.diff.expression_HAI$Description <- factor(gsea.diff.expression_HAI$Description, levels = gsea.diff.expression_HAI$Description[order(gsea.diff.expression_HAI$enrichmentScore)])

gsea.diff.expression_HAI %>%
    filter(p.adjust < 0.05) %>%
    ggplot(aes(enrichmentScore, Description, color = p.adjust, size = setSize)) +
    geom_point() +
    ggtitle("Biological Pathway (p.adj<0.05)") +
    scale_colour_gradient2(low = "blue", mid = "white", high = "red")

```

```{r}
target.genes <- toptags[[4]]$table
target.genes <- target.genes[target.genes$FDR < 0.05,]



gene.list <- target.genes$logFC
names(gene.list) <- rownames(target.genes)
gene.list <-  sort(gene.list, decreasing = TRUE)
```


Run GSEA
```{r}
ontlist <- c("BP", "CC", "MF")
simplist <- list()
resultlist <- list()
tib.list <- list()

for (i in 1:length(ontlist)){
  hsGO <- godata('org.Hs.eg.db', ont = ontlist[i])
  set.seed(86)
  gse <- gseGO(geneList=gene.list, 
             ont = ontlist[i], 
             keyType = "ENSEMBL", 
             minGSSize = 3, 
             maxGSSize = 800, 
             pvalueCutoff = 0.05, 
             verbose = TRUE, 
             OrgDb = get(organism), 
             pAdjustMethod = "none",
             seed=T)
  simp <- clusterProfiler::simplify(gse, 
                                    semData = hsGO,
                                    cutoff = 0.4) #change to reduce terms
  simplist[[i]] <- simp
  tib.list[[i]] <- as_tibble(simp@result)
}

gsea.diff.expression_sero <- as_tibble(rbind(tib.list[[1]], tib.list[[2]], tib.list[[3]]))

#write.csv(gsea.diff.expression_sero, file = "Seroconversion_GSEA.csv", row.names = TRUE)


gsea.diff.expression_sero$Description <- factor(gsea.diff.expression_sero$Description, levels = gsea.diff.expression_sero$Description[order(gsea.diff.expression_sero$enrichmentScore)])

gsea.diff.expression_sero %>%
    filter(p.adjust < 0.05) %>%
    ggplot(aes(enrichmentScore, Description, color = p.adjust, size = setSize)) +
    geom_point() +
    labs(x = "Enrichment Score",
         y = "Biological Pathway",
         title = "Pathway Enrichment: Seroconversion") +
    scale_colour_gradient2(name = "P Value") +
  scale_size(name = "Gene Set Size") +
    theme(text = element_text(size = 30)) +
    scale_y_discrete(labels = function(x) str_wrap(x, width = 40))



```

Plot all GSEA results in one plot
```{r}

gsea.diff.expression_HAI$group <- "HAI"
gsea.diff.expression_sero$group <- "Seroconversion"

gsea.diff.expression_all <- as_tibble(rbind(gsea.diff.expression_HAI, gsea.diff.expression_sero))

fig2$f <- gsea.diff.expression_all %>%
    filter(Description != "adaptive immune response based on somatic recombination of immune receptors built from immunoglobulin superfamily domains") %>%
  filter(Description != "biological_process") %>%
  filter(Description != "molecular_function") %>%
    ggplot(aes(enrichmentScore, Description, size = setSize)) +
    geom_point(aes(fill = p.adjust, size = setSize), colour="black", pch=21) +
  facet_grid(group~., scales = "free", space = "free")+
  labs(x = "Enrichment Score",
         y = "Description",
         title = "") +
  scale_size(name = "Gene Set Size", range = c(1, 20), limits = c(1, 100), breaks = c(5, 10, 50, 100)) +
  scale_fill_viridis_c(name = "P Value") +
  theme(text = element_text(size = 35, family = "Arial"),
          legend.text = element_text(size = 25),
          legend.key.height = unit(1, 'cm'),
          axis.text.y = element_text(size = 25))


pdf(paste0("fig2f.pdf"),
    height=12,
    width=18)
plot(fig2$f)
dev.off()

png(paste0("fig2f.png"),
    height=4800,
    width=6000, 
    units = "px",
    res = 300)
plot(fig2$f)
dev.off()
```

```{r}

fig2left <- ((fig2$b | fig2$c) / (fig2$d | fig2$e)) 


pdf(paste0("fig2bcde.pdf"),
    height=10,
    width=10)
plot(fig2left)
dev.off()
```