class2tree workaround

class2tree_workaround.R

library(taxize)
library(reshape2)

### stuff starts here

#### remove the rank filter from the clustering method

class2tree_helper_no_filter <- function(x){
  #x <- x[!x$rank == "no rank", ]
  df <- x[-nrow(x), 'id']
  names(df) <- x[-nrow(x), 'rank']
  df <- data.frame(t(data.frame(df)), stringsAsFactors = FALSE)
  data.frame(tip = x[nrow(x), "name"], df, stringsAsFactors = FALSE)
}


#### get classification & convert into dataframe

spnames <- c('Homo_sapiens',
             'Pan_troglodytes',
             'Macaca_mulatta',
             'Mus_musculus',
             'Rattus_norvegicus',
             'Bos_taurus',
             'Canis_lupus',
             'Ornithorhynchus_anatinus',
             'Xenopus_tropicalis',
             'Takifugu_rubripes',
             'Gallus_gallus',
             'Ciona_intestinalis',
             'Branchiostoma_floridae',
             'Schistosoma_mansoni',
             'Caenorhabditis_elegans',
             'Anopheles_gambiae',
             'Drosophila_melanogaster',
             'Ixodes_scapularis',
             'Ustilago_maydis',
             'Neurospora_crassa',
             'Monodelphis_domestica',
             'Danio_rerio',
             'Nematostella_vectensis',
             'Cryptococcus_neoformans')


out <- classification(spnames, db='ncbi')
df <- rbind.fill(lapply(out, class2tree_helper_no_filter))

# this is what class2tree produces on its own

tr <- class2tree(out)
plot(tr)

# this is the alternative that tries to take all levels, also unranked, into account



# initialize empty df for calculated distances

df_dist <- data.frame(x = character(nrow(df)*nrow(df)), y = character(nrow(df)*nrow(df)), dist = numeric(nrow(df)*nrow(df)), stringsAsFactors = FALSE)

# iterate over each row of classification DF & fill distance
# this is probably the most un-R thing ever but my tired brain didn't find any better solution
c = 1

for (r1 in 1:nrow(df)) {
  n1 = df[r1,1] 
  x = as.vector(t(df[r1,-1]))
  for (r2 in 1:nrow(df)) {
    n2 = df[r2,1]
    y = as.vector(t(df[r2,-1]))
    distance = length(union(x,y))-length(intersect(x,y))
    df_dist$x[c] = n1
    df_dist$y[c] = n2
    df_dist$dist[c] = distance
    c = c +1
  }
}

# convert long dataframe to distance object & plot hclust result
dist_matrix <- as.dist(acast(df_dist, x ~ y, value.var='dist', fun.aggregate = sum, margins=FALSE))
plot(hclust(dist_matrix))