My good friend Andrew recently posted this gist:
It shows how to solve a simple programming exercise know as
FizzBuzz in R, but also is clearly showing off his already amazing proficiency in the very new R package
magrittr. Those are some piping hot coding skillz (or should I say (not-a-)piping hot?)
Clearly this was a challenge to see who can come up with the most ridiculously long and complicated expression that uses only a single string of functions piped together with the magic of
%>%. In that spirit, I have decided to try and replicate one of the more complicated analyses and figures in my first paper—a phylogenetic ordination—using no intermediate variables and lots of
%>%-ing. Actually, I am going to use a different method for the phylogenetic ordination from the paper, and I think this one is better (I will post more on this later).
I download the data directly from the supplemental material on my paper, so you should be able to run this code on your computer if you want (as long as you have an internet connection). This requires the following packages:
library(ape) library(plyr) library(dplyr) library(magrittr) library(vegan) library(ggplot2) dat<-scan("http://www.plosone.org/article/fetchSingleRepresentation.action?uri=info:doi/10.1371/journal.pone.0007071.s002",what=character(0)) %>% #download data read.tree(text=.) %>% #make it a phylo object cophenetic %>% # turn it into a phylogenetic distance matrix l(.,x -> cmdscale(x,nrow(x)-1)) %>% # get phylogenetic principle components l(.,x -> x[order(rownames(x)), ]) %>% # order by rownames aaply("http://www.plosone.org/article/fetchSingleRepresentation.action?uri=info:doi/10.1371/journal.pone.0007071.s001" %>% read.csv %>% # read-in community data l(x -> x %>% set_rownames(paste(rownames(x),x[,2],sep="_"))) %>% #save treatment names for later in the rownames extract(c(-1,-2)) %>% # toss non-species data l(x -> x[,order(colnames(x))]) %>% # order by column names as.matrix, # make it a matrix 1,function(x, y) (y*x) %>% colSums, y=.) %>% #make a function to pipe in # community data as x, and phylo data (.) as y, then multiple each community abundance # by phylo principle components. now we have a phylogenetic feature vector for each # site! metaMDS("euclidean") %>% # use non-metric multidimensional scaling on phylogenetic features extract("points") %>% # pull out the ordinated points data.frame %>% # make it data.frame l(., x -> mutate(x,treatment=rownames(x) %>% strsplit("_") %>% laply(function(y) y))) %>% #extract treatment names from rownames ggplot(.,aes(x=points.MDS1,y=points.MDS2)) + geom_point(aes(color=factor(treatment)),size=5) #make a ggplot ## Run 0 stress 0.1055 ## Run 1 stress 0.1055 ## ... procrustes: rmse 0.01079 max resid 0.05315 ## Run 2 stress 0.1199 ## Run 3 stress 0.1198 ## Run 4 stress 0.1057 ## ... procrustes: rmse 0.006186 max resid 0.03488 ## Run 5 stress 0.1055 ## ... procrustes: rmse 0.01088 max resid 0.05316 ## Run 6 stress 0.121 ## Run 7 stress 0.1408 ## Run 8 stress 0.1055 ## ... procrustes: rmse 0.01081 max resid 0.05317 ## Run 9 stress 0.1055 ## ... New best solution ## ... procrustes: rmse 0.000166 max resid 0.0007816 ## *** Solution reached dat #plot ordination!!
You can also find the above as a gist.
We can see that the main difference between the disturbed and undisturbed sites is that there is a much larger variance in phylogenetic composition between disturbed sites. This has interesting implications which has inspired me to do a follow-up to my old paper, on
this blog. Look for that soon.
So that's a phylogenetic ordination, done in a single line of code (well, it would be a single line if I removed all the linebreaks). Not a single intermediate variable was used. Note that if you do try this code, you will get some warnings, they don't affect the outcome. It's not nearly as elegant as Andrew's, but I would hate to see what this would look like if I tried nesting all of these functions!
Did I mention that I love
magrittr (you could probably guess that from the gif that I made, below).