Here is a minimal example but @cmartino can you please advise on the most appropriate way of selecting the taxa and for scaling their lengths? I am not 100% what I did is valid.
library(tidyverse)
library(qiime2R)
ord<-read_qza("~/Downloads/ordination.qza")
meta<-read_tsv(file = "~/Downloads/metadata.txt") %>%
rename(SampleID=`#SampleID`) %>%
filter(SampleID!="#q2:types")
tax<-read_qza("~/Downloads/taxonomy.qza")$data %>%
rename(FeatureID=Feature.ID)
#create the base plot with only the arrows
baseplot<-
ggplot() +
theme_bw() +
xlab(paste(round(100*ord$data$ProportionExplained[1],2),"%")) +
ylab(paste(round(100*ord$data$ProportionExplained[2],2),"%")) +
ggtitle("DEICODE biplot of Moving Pictures Data") +
geom_segment(data=ord$data$Species %>%
mutate(a=sqrt(PC1^2+PC2^2)) %>% # calculate the distance from the origin
top_n(8, a) %>% #keep 8 furthest away points
mutate(PC1=PC1*0.3, PC2=PC2*0.3) %>% # scale arrows linearly... is this ok?
left_join(tax),
aes(x=0, xend=PC1, y=0, yend=PC2, color=Taxon),
arrow = arrow(length = unit(0.3,"cm"))
)
# now overlay samples
baseplot +
geom_point(
data=ord$data$Vectors %>%
left_join(meta),
aes(x=PC1, y=PC2, fill=BodySite),
shape=21
)
