-## dist.topo.R (2008-02-27)
+## dist.topo.R (2010-05-25)
## Topological Distances, Tree Bipartitions,
## Consensus Trees, and Bootstrapping Phylogenies
-## Copyright 2005-2008 Emmanuel Paradis
+## Copyright 2005-2010 Emmanuel Paradis
## This file is part of the R-package `ape'.
## See the file ../COPYING for licensing issues.
dist.topo <- function(x, y, method = "PH85")
{
- if (method == "BHV01" && (is.null(x$edge.length) || is.null(y$edge.length)))
- stop("trees must have branch lengths for Billera et al.'s distance.")
- n <- length(x$tip.label)
- bp1 <- .Call("bipartition", x$edge, n, x$Nnode, PACKAGE = "ape")
+ if (method == "score" && (is.null(x$edge.length) || is.null(y$edge.length)))
+ stop("trees must have branch lengths for Billera et al.'s distance.")
+ nx <- length(x$tip.label)
+ x <- unroot(x)
+ y <- unroot(y)
+ bp1 <- .Call("bipartition", x$edge, nx, x$Nnode, PACKAGE = "ape")
bp1 <- lapply(bp1, function(xx) sort(x$tip.label[xx]))
- bp2 <- .Call("bipartition", y$edge, n, y$Nnode, PACKAGE = "ape")
- bp2 <- lapply(bp2, function(xx) sort(y$tip.label[xx]))
+ ny <- length(y$tip.label) # fix by Otto Cordero
+ ## fix by Tim Wallstrom:
+ bp2.tmp <- .Call("bipartition", y$edge, ny, y$Nnode, PACKAGE = "ape")
+ bp2 <- lapply(bp2.tmp, function(xx) sort(y$tip.label[xx]))
+ bp2.comp <- lapply(bp2.tmp, function(xx) setdiff(1:ny, xx))
+ bp2.comp <- lapply(bp2.comp, function(xx) sort(y$tip.label[xx]))
+ ## End
q1 <- length(bp1)
q2 <- length(bp2)
if (method == "PH85") {
p <- 0
for (i in 1:q1) {
for (j in 1:q2) {
- if (identical(all.equal(bp1[[i]], bp2[[j]]), TRUE)) {
+ if (identical(bp1[[i]], bp2[[j]]) | identical(bp1[[i]], bp2.comp[[j]])) {
p <- p + 1
break
}
}
}
- dT <- if (q1 == q2) 2*(q1 - p) else 2*(min(q1, q2) - p) + abs(q1 - q2)
+ dT <- q1 + q2 - 2 * p # same than:
+ ##dT <- if (q1 == q2) 2*(q1 - p) else 2*(min(q1, q2) - p) + abs(q1 - q2)
}
- if (method == "BHV01") {
+ if (method == "score") {
dT <- 0
found1 <- FALSE
found2 <- logical(q2)
found2[1] <- TRUE
for (i in 2:q1) {
for (j in 2:q2) {
- if (identical(bp1[[i]], bp2[[j]])) {
- dT <- dT + abs(x$edge.length[which(x$edge[, 2] == n + i)] -
- y$edge.length[which(y$edge[, 2] == n + j)])
+ if (identical(bp1[[i]], bp2[[j]]) | identical(bp1[[i]], bp2.comp[[j]])) {
+ dT <- dT + (x$edge.length[which(x$edge[, 2] == nx + i)] -
+ y$edge.length[which(y$edge[, 2] == ny + j)])^2
found1 <- found2[j] <- TRUE
break
}
}
if (found1) found1 <- FALSE
- else dT <- dT + x$edge.length[which(x$edge[, 2] == n + i)]
+ else dT <- dT + (x$edge.length[which(x$edge[, 2] == nx + i)])^2
}
if (!all(found2))
- dT <- dT + sum(y$edge.length[y$edge[, 2] %in% (n + which(!found2))])
+ dT <- dT + sum((y$edge.length[y$edge[, 2] %in% (ny + which(!found2))])^2)
+ dT <- sqrt(dT)
}
dT
}
-prop.part <- function(..., check.labels = FALSE)
+.compressTipLabel <- function(x)
+{
+ ## 'x' is a list of objects of class "phylo" possibly with no class
+ if (!is.null(attr(x, "TipLabel"))) return(x)
+ ref <- x[[1]]$tip.label
+ if (any(table(ref) != 1))
+ stop("some tip labels are duplicated in tree no. 1")
+ n <- length(ref)
+ for (i in 2:length(x)) {
+ label <- x[[i]]$tip.label
+ if (!identical(label, ref)) {
+ if (length(label) != length(ref))
+ stop(paste("tree no.", i, "has a different number of tips"))
+ ilab <- match(label, ref)
+ ## can use tabulate here because 'ilab' contains integers
+ if (any(is.na(ilab)))
+ stop(paste("tree no.", i, "has different tip labels"))
+### <FIXME> the test below does not seem useful anymore
+### if (any(tabulate(ilab) > 1))
+### stop(paste("some tip labels are duplicated in tree no.", i))
+### </FIXME>
+ ie <- match(1:n, x[[i]]$edge[, 2])
+ x[[i]]$edge[ie, 2] <- ilab
+ }
+ x[[i]]$tip.label <- NULL
+ }
+ x[[1]]$tip.label <- NULL
+ attr(x, "TipLabel") <- ref
+ x
+}
+
+prop.part <- function(..., check.labels = TRUE)
{
obj <- list(...)
if (length(obj) == 1 && class(obj[[1]]) != "phylo")
- obj <- unlist(obj, recursive = FALSE)
+ obj <- obj[[1]]
+ ## <FIXME>
+ ## class(obj) <- NULL # needed?
+ ## </FIXME>
ntree <- length(obj)
- if (!check.labels) {
- for (i in 1:ntree) storage.mode(obj[[i]]$Nnode) <- "integer"
- clades <- .Call("prop_part", obj, ntree, TRUE, PACKAGE = "ape")
- attr(clades, "number") <- attr(clades, "number")[1:length(clades)]
- attr(clades, "labels") <- obj[[1]]$tip.label
- } else {
- bp <- .Call("bipartition", obj[[1]]$edge, length(obj[[1]]$tip.label),
- obj[[1]]$Nnode, PACKAGE = "ape")
- clades <- lapply(bp, function(xx) sort(obj[[1]]$tip.label[xx]))
- no <- rep(1, length(clades))
-
- if (ntree > 1) {
- for (k in 2:ntree) {
- bp <- .Call("bipartition", obj[[k]]$edge,
- length(obj[[k]]$tip.label), obj[[k]]$Nnode,
- PACKAGE = "ape")
- bp <- lapply(bp, function(xx) sort(obj[[k]]$tip.label[xx]))
- for (i in 1:length(bp)) {
- done <- FALSE
- for (j in 1:length(clades)) {
- if (identical(all.equal(bp[[i]], clades[[j]]), TRUE)) {
- no[j] <- no[j] + 1
- done <- TRUE
- break
- }
- }
- if (!done) {
- clades <- c(clades, bp[i])
- no <- c(no, 1)
- }
- }
- }
- }
- attr(clades, "number") <- no
- }
+ if (ntree == 1) check.labels <- FALSE
+ if (check.labels) obj <- .compressTipLabel(obj)
+ for (i in 1:ntree) storage.mode(obj[[i]]$Nnode) <- "integer"
+ ## <FIXME>
+ ## The 1st must have tip labels
+ ## Maybe simply pass the number of tips to the C code??
+ if (!is.null(attr(obj, "TipLabel")))
+ for (i in 1:ntree) obj[[i]]$tip.label <- attr(obj, "TipLabel")
+ ## </FIXME>
+ clades <- .Call("prop_part", obj, ntree, TRUE, PACKAGE = "ape")
+ attr(clades, "number") <- attr(clades, "number")[1:length(clades)]
+ attr(clades, "labels") <- obj[[1]]$tip.label
class(clades) <- "prop.part"
clades
}
n
}
-boot.phylo <- function(phy, x, FUN, B = 100, block = 1)
+boot.phylo <- function(phy, x, FUN, B = 100, block = 1, trees = FALSE)
{
- if (is.list(x)) {
- if (class(x) == "DNAbin") x <- as.matrix(x)
+ if (is.list(x) && !is.data.frame(x)) {
+ if (inherits(x, "DNAbin")) x <- as.matrix(x)
else {
nm <- names(x)
n <- length(x)
}
for (i in 1:B) storage.mode(boot.tree[[i]]$Nnode) <- "integer"
storage.mode(phy$Nnode) <- "integer"
- attr(.Call("prop_part", c(list(phy), boot.tree), B + 1, FALSE,
- PACKAGE = "ape"), "number") - 1
+ ans <- attr(.Call("prop_part", c(list(phy), boot.tree),
+ B + 1, FALSE, PACKAGE = "ape"), "number") - 1
+ if (trees) ans <- list(BP = ans, trees = boot.tree)
+ ans
}
-consensus <- function(..., p = 1)
+consensus <- function(..., p = 1, check.labels = TRUE)
{
+ foo <- function(ic, node) {
+ ## ic: index of 'pp'
+ ## node: node number in the final tree
+ pool <- pp[[ic]]
+ if (ic < m) {
+ for (j in (ic + 1):m) {
+ wh <- match(pp[[j]], pool)
+ if (!any(is.na(wh))) {
+ edge[pos, 1] <<- node
+ pool <- pool[-wh]
+ edge[pos, 2] <<- nextnode <<- nextnode + 1L
+ pos <<- pos + 1L
+ foo(j, nextnode)
+ }
+ }
+ }
+ size <- length(pool)
+ if (size) {
+ ind <- pos:(pos + size - 1)
+ edge[ind, 1] <<- node
+ edge[ind, 2] <<- pool
+ pos <<- pos + size
+ }
+ }
obj <- list(...)
- if (length(obj) == 1 && class(obj[[1]]) != "phylo")
- obj <- unlist(obj, recursive = FALSE)
+ if (length(obj) == 1) {
+ ## better than unlist(obj, recursive = FALSE)
+ ## because "[[" keeps the class of 'obj':
+ obj <- obj[[1]]
+ if (class(obj) == "phylo") return(obj)
+ }
+ if (!is.null(attr(obj, "TipLabel")))
+ labels <- attr(obj, "TipLabel")
+ else {
+ labels <- obj[[1]]$tip.label
+ if (check.labels) obj <- .compressTipLabel(obj)
+ }
ntree <- length(obj)
## Get all observed partitions and their frequencies:
- pp <- prop.part(obj, check.labels = TRUE)
+ pp <- prop.part(obj, check.labels = FALSE)
## Drop the partitions whose frequency is less than 'p':
pp <- pp[attr(pp, "number") >= p * ntree]
## Get the order of the remaining partitions by decreasing size:
- ind <- rev(sort(unlist(lapply(pp, length)),
- index.return = TRUE)$ix)
- pp <- lapply(pp, function(xx) paste("IMPROBABLE_PREFIX", xx,
- "IMPROBABLE_SUFFIX", sep = "_"))
- STRING <- paste(pp[[1]], collapse = ",")
- STRING <- paste("(", STRING, ");", sep = "")
- for (i in ind[-1]) {
- ## 1. Delete all tips in the focus partition:
- STRING <- unlist(strsplit(STRING, paste(pp[[i]], collapse = "|")))
- ## 2. Put the partition in any of the created gaps:
- STRING <- c(STRING[1],
- paste("(", paste(pp[[i]], collapse = ","), ")", sep = ""),
- STRING[-1])
- ## 3. Stick back the Newick string:
- STRING <- paste(STRING, collapse = "")
+ ind <- sort(unlist(lapply(pp, length)), decreasing = TRUE,
+ index.return = TRUE)$ix
+ pp <- pp[ind]
+ n <- length(labels)
+ m <- length(pp)
+ edge <- matrix(0L, n + m - 1, 2)
+ if (m == 1) {
+ edge[, 1] <- n + 1L
+ edge[, 2] <- 1:n
+ } else {
+ nextnode <- n + 1L
+ pos <- 1L
+ foo(1, nextnode)
}
- ## Remove the extra commas:
- STRING <- gsub(",{2,}", ",", STRING)
- STRING <- gsub("\\(,", "\\(", STRING)
- STRING <- gsub(",\\)", "\\)", STRING)
- STRING <- gsub("IMPROBABLE_PREFIX_", "", STRING)
- STRING <- gsub("_IMPROBABLE_SUFFIX", "", STRING)
- read.tree(text = STRING)
+ structure(list(edge = edge, tip.label = labels,
+ Nnode = m), class = "phylo")
}