3.P259.pDC_GSEA.Rmd

---
title: 'P259: Gene set enrichment analysis (GSEA)'
subtitle: "Dendritic cells (pDC)"
author: "Kim Dill-McFarland, kadm@uw.edu"
date: "version `r format(Sys.time(), '%B %d, %Y')`"
output:
  html_document:
    toc: yes
    toc_depth: 3
    toc_float:
      collapsed: no
  pdf_document:
    toc: yes
    toc_depth: '3'
editor_options:
  chunk_output_type: console
---
# Background

The purpose of this workflow is to perform GSEA for the impacts of human rhinovirus (RV) infection, eosinophil (EOS) supernatant, and Anti-IL5 therapy.

# Setup
Load packages

```{r message=FALSE, warning=FALSE}
# Data manipulation and figures
library(tidyverse)
# Multipanel figures
library(cowplot)

#GSEA
library(fgsea)
library(gage)

#Print pretty tables to Rmd
library(knitr)
library(kableExtra)
```

Set seed

```{r}
set.seed(589)
```

# Load data
## RNA-seq

Contrast model results.

```{r results.data, message=FALSE}
pval_2 <- read_csv("results/gene_level/P259.2_gene_pval.csv") %>% 
  filter(model=="contrasts")
pval_1 <- read_csv("results/gene_level/P259.1_gene_pval.csv") %>% 
  filter(model=="contrasts")
```

Extract and format fold change (FC) for each contrast.

```{r}
gene.ls <- list()

for (contrast in unique(pval_2$group)){
  #Subset to contrast of interest
  pval.temp <- pval_2 %>% filter(group == contrast)
  
  genes.temp <- pval.temp$logFC
  names(genes.temp) <- pval.temp$hgnc_symbol
  
  list.name <- paste(gsub(" - ", ".", contrast), 2, sep=".")
  gene.ls[[list.name]] <- genes.temp
}

for (contrast in unique(pval_1$group)){
  #Subset to contrast of interest
  pval.temp <- pval_1 %>% filter(group == contrast)
  
  genes.temp <- pval.temp$logFC
  names(genes.temp) <- pval.temp$hgnc_symbol
  
  list.name <- paste(gsub(" - ", ".", contrast), 1, sep=".")
  gene.ls[[list.name]] <- genes.temp
}
```

## Broad gene sets

From <https://www.gsea-msigdb.org/gsea/msigdb/collections.jsp>. Downloaded in `data_clean/Broad_gmt/`

# Gene set enrichment analysis (GSEA)

The following function performs GSEA using both fast gene set enrichment analysis (`fgsea`) and generally applicable gene-set enrichment (`gage`). 

```{r}
source("https://raw.githubusercontent.com/kdillmcfarland/R_bioinformatic_scripts/master/GSEA_fxn.R")
```

## Run GSEA
#### Hallmark (H)

```{r h, warnings=FALSE, message=FALSE, eval=FALSE}
GSEA(gene_list = gene.ls, 
     gmt_file="data_clean/Broad_gmt/h.all.v7.4.symbols.gmt")
```

#### Curated canonical pathway gene sets (C2:CP)

```{r c2, warnings=FALSE, message=FALSE, eval=FALSE}
GSEA(gene_list = gene.ls, 
     gmt_file="data_clean/Broad_gmt/c2.cp.v7.4.symbols.gmt")
```

#### GO biological procress gene sets (C5:GO:BP)

```{r c5, warnings=FALSE, message=FALSE, eval=FALSE}
GSEA(gene_list = gene.ls, 
     gmt_file="data_clean/Broad_gmt/c5.go.bp.v7.4.symbols.gmt")
```

## Significant GSEA

```{r echo=FALSE, message=FALSE}
fdr.cut <- 0.1
#Load results for faster knitting
h_GSEA.result <- read_csv("results/GSEA/h_GSEA.result.csv")
c2_GSEA.result <- read_csv("results/GSEA/c2_GSEA.result.csv")
c5_GSEA.result <- read_csv("results/GSEA/c5_GSEA.result.csv")

#Combine and filter signif
GSEA.all <- bind_rows(h_GSEA.result,c2_GSEA.result,c5_GSEA.result)
#list all terms with at least 1 significant treatment contrast
term.OI.t <- GSEA.all %>% 
  filter(fgsea.FDR <= fdr.cut & gage.FDR <= fdr.cut) %>% 
  filter(group %in% c("AntiIL5_nonenone_none", "AntiIL5_HRVnone_HRV",
                      "EOSsupp_nonenone_none", "EOSsupp_HRVnone_HRV")) %>% 
  distinct(pathway) %>% unlist(use.names=FALSE)

#list all terms with at least 1 significant virus contrast 
term.OI.v <- GSEA.all %>% 
  filter(fgsea.FDR <= fdr.cut & gage.FDR <= fdr.cut) %>% 
  filter(group %in% c("none_HRVnone_none","AntiIL5_HRVAntiIL5_none",
                      "EOSsupp_HRVEOSsupp_none")) %>% 
  distinct(pathway) %>% unlist(use.names=FALSE)
   
#Find terms with both t and v cutoffs
term.OI <- intersect(term.OI.t, term.OI.v)

GSEA.signif <- GSEA.all %>% 
  filter(pathway %in% term.OI)
```

Gene sets of interest are those significant for both virus AND EOS supernatant or Anti-IL5 therapy. Results are only considered significant if both fgsea and gage methods meet the FDR threshold in the same fold change direction.

```{R echo=FALSE}
GSEA.signif %>% 
  select(pathway, group, fgsea.FDR,fgsea.NES) %>% 
  separate(pathway, into=c("set","pathway"), sep="_", extra = "merge") %>% 
  select(set, pathway, group, fgsea.FDR,fgsea.NES) %>% 
  arrange(set, pathway, fgsea.NES) %>% 
  kable() %>%
kable_styling(bootstrap_options = "striped", full_width = FALSE) %>% 
  collapse_rows(1:2, valign = "top")
```

# Hypergeometric enrichment of DEGs

```{r echo=FALSE}
fdr.cut <- 0.1
```

Test for Broad gene set enrichment in differentially expressed genes (DEG) as defined by change with virus (*e.g.* significant for virus in untreated and/or treated donors) AND different between untreated and treated donors in media OR RV. DEGs are defined at FDR < `r fdr.cut`

```{r}
#Define change with virus
DEG1.v <- read_csv("results/gene_level/P259.1_gene_pval.csv") %>% 
  filter(group %in% c("none_HRV - none_none", "EOS.supp_HRV - EOS.supp_none") &
           adj.P.Val <= fdr.cut)
#Define change with treament
DEG1.t <- read_csv("results/gene_level/P259.1_gene_pval.csv") %>% 
  filter(group %in% c("EOS.supp_none - none_none", "EOS.supp_HRV - none_HRV") &
           adj.P.Val <= fdr.cut)

#Define change with virus
DEG2.v <- read_csv("results/gene_level/P259.2_gene_pval.csv") %>% 
  filter(group %in% c("none_HRV - none_none", "AntiIL5_HRV - AntiIL5_none") &
           adj.P.Val <= fdr.cut)
#Define change with treament
DEG2.t <- read_csv("results/gene_level/P259.2_gene_pval.csv") %>% 
  filter(group %in% c("AntiIL5_none - none_none", "AntiIL5_HRV - none_HRV") &
           adj.P.Val <= fdr.cut)

#Format in list
DEG.ls <- list()
DEG.ls[["EOSsup"]] <- intersect(DEG1.v$hgnc_symbol, DEG1.t$hgnc_symbol)
DEG.ls[["AntiIL5"]] <- intersect(DEG2.v$hgnc_symbol, DEG2.t$hgnc_symbol)
```

```{r message=FALSE, warning=FALSE}
#Script for running term enrichment
source("https://raw.githubusercontent.com/kdillmcfarland/R_bioinformatic_scripts/master/hypergeo_enricher.R")
```

## Run enrichment
#### Hallmark (H)

```{r h.2, warnings=FALSE, message=FALSE, eval=FALSE}
enrich.fxn(gene.list = DEG.ls, ID.type="SYMBOL",
           category = "H",
           genome = "org.Hs.eg.db", 
           basename = "DEG",
           outdir = "results/enrichment/")
```

#### Curated canonical pathway gene sets (C2:CP)

```{r c2.2, warnings=FALSE, message=FALSE, eval=FALSE}
enrich.fxn(gene.list = DEG.ls, ID.type="SYMBOL",
           category = "C2", subcategory="CP",
           genome = "org.Hs.eg.db", 
           basename = "DEG",
           outdir = "results/enrichment/")
```

#### GO biological procress gene sets (C5:GO:BP)

```{r c5.2, warnings=FALSE, message=FALSE, eval=FALSE}
enrich.fxn(gene.list = DEG.ls, ID.type="SYMBOL",
           category = "C5", subcategory = "GO:BP",
           genome = "org.Hs.eg.db", 
           basename = "DEG",
           outdir = "results/enrichment/")
```

## Significant enrichment

```{r echo=FALSE, message=FALSE}
fdr.cut <- 0.05
#Load results for faster knitting
h_enrich.result <- read_csv("results/enrichment/enrich_DEG_H.csv")
c2_enrich.result <- read_csv("results/enrichment/enrich_DEG_C2_CP.csv")
c5_enrich.result <- read_csv("results/enrichment/enrich_DEG_C5_GO.BP.csv")

#Combine and filter signif
enrich.signif <- bind_rows(h_enrich.result,c2_enrich.result,c5_enrich.result)%>% 
  filter(p.adjust <= fdr.cut) 
```

Gene sets enriched in DEGs at FDR < `r fdr.cut`

```{R echo=FALSE}
enrich.signif %>% 
  separate(Description, into=c("set","pathway"), sep="_", extra = "merge") %>% 
  dplyr::select(group, set, pathway, p.adjust, SYMBOLs) %>% 
  arrange(group, set, p.adjust) %>% 
  kable() %>%
kable_styling(bootstrap_options = "striped", full_width = FALSE) %>% 
  collapse_rows(1:2, valign = "top")
```

# R session

```{r}
sessionInfo()
```

***