自己说好不是好，用户说好才是真的好！

ETE是python里功能最强大的进化树可视化软件，然而有python用户最终选择了ggtree，我只能默默地点赞！这是一篇很好的入门介绍ggtree的文章，因为它一步步地添加图层并给出详细的解释。

Experiments with ggtree

Being predominantly a python user when it comes to bioinformatics analyses I was keen to try and find a python library to draw trees with associated ‘heatmaps’ that display categorical meta data alongside the tip labels of the tree. I really tried to like the ete3 library that some had recommended but found the interface to the functions too obtuse. Other colleagues suggested the ggtree package in R. Although it has a similar API to ggplot2, which in itself takes some getting used to, the documentation is good and with a little effort taken to read the extensive docs has a fairly intuitive interface.

This post will explore my explorations in how phylogenies and associated heatmaps can be drawn using the ggtree package in R. I would strongly recommend using RStudio since it makes working within R a breeze, particularly with graphical outputs.

• Reading in a tree

   library("ape")  library("ggtree")  tree <- read.tree("/path/to/newick_file")

• Plot the tree

   p <- ggtree(tree)  plot(p)

• I prefer a right ladderized tree

   p <- ggtree(tree, right = TRUE)  plot(p)

• Let’s add some tip lables and a title

   p <- ggtree(tree, right = TRUE) + ggtitle("Test Tree") + geom_tiplab(size = 2)  plot(p)

• If you prefer right aligned labels this can be done, and we’ll also add a scale bar. N.B Addition of ggplot2::xlim(0, 0.3) is necessary to stop truncated labels when aligned right, the second parameter needs to be determined by trial and error. See FAQ

   p <- ggtree(tree, right = TRUE) + ggtitle("Test Tree") + geom_tiplab(size = 2, align=TRUE, linesize=.25)  + geom_treescale(x=0.05, y=0, offset=2, fontsize = 3) + ggplot2::xlim(0, 0.3) plot(p)

• Adding a bootstrap value is a bit more fun, requiring some data manipulation.

• Get all data from the tree and find only non leaf nodes.

• Convert the lables to numeric values and only keep those where the value is >65

• Add it to the tree as geom_text. The hjust and vjust parameters can be eited to control position (again by trial and error)

• N.B This is also found in the FAQ

• Now the tree is looking kinda OK, we can get round to adding the heatmap.

  Data is provided in the format as a tsv and read using the following code. Header and row names are specified using the two parameters header=TRUE and row.names=1. The check.names=FALSE param is necessary in case sample names begin with a numeric.

  meta_data <- read.table("meta.tsv", sep="\t", header=TRUE,check.names=FALSE, stringsAsFactor=F, row.names = 1)

  sample	phenotype	MIC
  sample_60546	resistant	10
  sample_40537	high level resistant	256
  sample_00125	sensitive	0
  sample_01454	intermediate	0.5

• We can plot a heatmap of the data contained in this file with the following code

  hm <- gheatmap(p,meta_data, offset = 0.02, width=0.15, font.size=3, colnames_position= "top", colnames_angle = 90, colnames_offset_y = 0, hjust = 0) + scale_fill_manual(values=c("sensitive" = "green", "intermediate" = "turquoise", "resistant" = "blue", "high level resistant" = "purple3", "0" = "white", "0.25" = "white", "0.5" = "gold", "10" = "darkorange2", "15" = "darkorange2", "20" = "darkorange2", "256" = "firebrick3"))plot(hm)

  Breaking this down:

• The offset param determines the distance between the tree and the heatmap (trial and error to set the best distance)

• width represents the proportion of the entire plot that will be used by the heatmap

• font.size is the size of the column headings

• colnames_position is the position of the columns labels (could also be ‘bottom’)

• colnames_angle is self explanatory

• colnames_offset_y and hjust allow fine tuning of the column name position

• scale_fill_manual - this is the most critical param in which the pairs in the values vector of the format 

  = determine which values in the meta data tsv file will be coloured with which colour (see [R colour chart](http://www.stat.columbia.edu/~tzheng/files/Rcolor.pdf))

• The legend may not be broken up appropriately into value groups so I would suggest a bit of manipulation in a vecor drawing program such as Inkscape to get it publication ready

赞赏