---
title: "Apply dendrogram labels (in cummeRbund)"
author: "Matthew L Bendall"
date: "September 12, 2016"
output: github_document
always_allow_html: yes
---

```{r setup, include=FALSE}
library(cummeRbund)
library(knitr)
library(DT)

options(digits=6)
knit_hooks$set(inline = function(x) {prettyNum(x, big.mark=",")})
knitr::opts_chunk$set(echo = TRUE)
```

### Load cummeRbund data

```{r load-data}
# Use CummeRbund to load data from cuffdiff output
# This takes awhile the first time around, but is faster on subsequent loads
cuff <- readCufflinks('/Users/bendall/Projects/asthma/results20160218/cuffdiff',
                      genome="hg38")
cuff
```

### Take a look at the replicate-level data

```{r}
DT::datatable(replicates(cuff),
              options=list(searching=FALSE, scrollX=TRUE))
```

### Plot the dendrogram

```{r}
dend.rep <- csDendro(genes(cuff), replicates=T)
```

As you can see here, the tips of the resulting dendrogram have useless names.
Plus, these names are not stable across analyses with different groupings or
numbers of samples.

### Set up vector with desired names

Create a named vector where the values are the desired names, and the names are
the labels used in the above dendrogram. In this case, the files used in the
analysis were named with the patient ID, so we can extract the patient IDs from
the `file` column.

```{r}
# goodnames (in the same order as the table):
goodnames <- gsub('/lustre/groups/cbi/asthma/host/hg38/', '', replicates(cuff)$file)
goodnames <- gsub('/abundances.cxb', '', goodnames)

# give the vector names using the rep_name column
names(goodnames) <- replicates(cuff)$rep_name
goodnames
```

Check that the new vector matches the `replicates(cuff)` table.

### Relabel dendrogram

Set the labels in the new dendrogram to desired labels. The `labels` function
is in the `dendextend` package (install using `install.packages("dendextend")`)

```{r}
newdend <- dend.rep
dendextend::labels(newdend) <- goodnames[labels(newdend)]
plot(newdend)
```

### Alternate approach using ggdendro

You can make more elaborate plots using the ggdendro package. This allows you to
work with the dendrogram like a ggplot object.

```{r}
library(ggplot2)
library(ggdendro) # Loads `ggendro::dendro_data` as `dendro_data`

# Convert to dendro data object
ggd <- ggdendro::dendro_data(dend.rep)
# Set up data.frame with the other variables
dvars <- replicates(cuff)
rownames(dvars) <- dvars$rep_name
# Reorder data.frame (same order as ggd)
dvars <- dvars[as.character(ggd$labels$label),]
# Add the good names to data.frame
dvars$goodname <- goodnames[rownames(dvars)]

ggdendrogram(ggd, rotate=T) +
        geom_text(data=ggd$labels, aes(x, y-0.05, label=dvars$goodname, color=dvars$sample_name)) +
        coord_flip() + scale_y_reverse() + theme_dendro() + theme(legend.position="none")

```