1 ## dist.topo.R (2011-03-26)
3 ## Topological Distances, Tree Bipartitions,
4 ## Consensus Trees, and Bootstrapping Phylogenies
6 ## Copyright 2005-2011 Emmanuel Paradis
8 ## This file is part of the R-package `ape'.
9 ## See the file ../COPYING for licensing issues.
11 dist.topo <- function(x, y, method = "PH85")
13 if (method == "score" && (is.null(x$edge.length) || is.null(y$edge.length)))
14 stop("trees must have branch lengths for branch score distance.")
15 nx <- length(x$tip.label)
18 bp1 <- .Call("bipartition", x$edge, nx, x$Nnode, PACKAGE = "ape")
19 bp1 <- lapply(bp1, function(xx) sort(x$tip.label[xx]))
20 ny <- length(y$tip.label) # fix by Otto Cordero
21 ## fix by Tim Wallstrom:
22 bp2.tmp <- .Call("bipartition", y$edge, ny, y$Nnode, PACKAGE = "ape")
23 bp2 <- lapply(bp2.tmp, function(xx) sort(y$tip.label[xx]))
24 bp2.comp <- lapply(bp2.tmp, function(xx) setdiff(1:ny, xx))
25 bp2.comp <- lapply(bp2.comp, function(xx) sort(y$tip.label[xx]))
29 if (method == "PH85") {
33 if (identical(bp1[[i]], bp2[[j]]) | identical(bp1[[i]], bp2.comp[[j]])) {
39 dT <- q1 + q2 - 2 * p # same than:
40 ##dT <- if (q1 == q2) 2*(q1 - p) else 2*(min(q1, q2) - p) + abs(q1 - q2)
42 if (method == "score") {
49 if (identical(bp1[[i]], bp2[[j]]) | identical(bp1[[i]], bp2.comp[[j]])) {
50 dT <- dT + (x$edge.length[which(x$edge[, 2] == nx + i)] -
51 y$edge.length[which(y$edge[, 2] == ny + j)])^2
52 found1 <- found2[j] <- TRUE
56 if (found1) found1 <- FALSE
57 else dT <- dT + (x$edge.length[which(x$edge[, 2] == nx + i)])^2
60 dT <- dT + sum((y$edge.length[y$edge[, 2] %in% (ny + which(!found2))])^2)
66 .compressTipLabel <- function(x)
68 ## 'x' is a list of objects of class "phylo" possibly with no class
69 if (!is.null(attr(x, "TipLabel"))) return(x)
70 ref <- x[[1]]$tip.label
71 if (any(table(ref) != 1))
72 stop("some tip labels are duplicated in tree no. 1")
77 label <- x[[i]]$tip.label
78 if (!identical(label, ref)) {
79 if (length(label) != length(ref))
80 stop(paste("tree no.", i, "has a different number of tips"))
81 ilab <- match(label, ref)
82 ## can use tabulate here because 'ilab' contains integers
84 stop(paste("tree no.", i, "has different tip labels"))
85 ### <FIXME> the test below does not seem useful anymore
86 ### if (any(tabulate(ilab) > 1))
87 ### stop(paste("some tip labels are duplicated in tree no.", i))
89 ie <- match(1:n, x[[i]]$edge[, 2])
90 x[[i]]$edge[ie, 2] <- ilab
92 x[[i]]$tip.label <- NULL
95 x[[1]]$tip.label <- NULL
96 attr(x, "TipLabel") <- ref
100 prop.part <- function(..., check.labels = TRUE)
103 if (length(obj) == 1 && class(obj[[1]]) != "phylo")
106 ## class(obj) <- NULL # needed? apparently not, see below (2010-11-18)
109 if (ntree == 1) check.labels <- FALSE
110 if (check.labels) obj <- .compressTipLabel(obj) # fix by Klaus Schliep (2011-02-21)
111 for (i in 1:ntree) storage.mode(obj[[i]]$Nnode) <- "integer"
113 ## The 1st must have tip labels
114 ## Maybe simply pass the number of tips to the C code??
115 obj <- .uncompressTipLabel(obj) # fix a bug (2010-11-18)
117 clades <- .Call("prop_part", obj, ntree, TRUE, PACKAGE = "ape")
118 attr(clades, "number") <- attr(clades, "number")[1:length(clades)]
119 attr(clades, "labels") <- obj[[1]]$tip.label
120 class(clades) <- "prop.part"
124 print.prop.part <- function(x, ...)
126 if (is.null(attr(x, "labels"))) {
127 for (i in 1:length(x)) {
128 cat("==>", attr(x, "number")[i], "time(s):")
129 print(x[[i]], quote = FALSE)
132 for (i in 1:length(attr(x, "labels")))
133 cat(i, ": ", attr(x, "labels")[i], "\n", sep = "")
135 for (i in 1:length(x)) {
136 cat("==>", attr(x, "number")[i], "time(s):")
137 print(x[[i]], quote = FALSE)
142 summary.prop.part <- function(object, ...) attr(object, "number")
144 plot.prop.part <- function(x, barcol = "blue", leftmar = 4, ...)
146 if (is.null(attr(x, "labels")))
147 stop("cannot plot this partition object; see ?prop.part for details.")
149 n <- length(attr(x, "labels"))
150 layout(matrix(1:2, 2, 1), heights = c(1, 3))
151 par(mar = c(0.1, leftmar, 0.1, 0.1))
152 plot(1:L, attr(x, "number"), type = "h", col = barcol, xlim = c(1, L),
153 xlab = "", ylab = "Frequency", xaxt = "n", bty = "n")
154 plot(0, type = "n", xlim = c(1, L), ylim = c(1, n),
155 xlab = "", ylab = "", xaxt = "n", yaxt = "n")
156 for (i in 1:L) points(rep(i, length(x[[i]])), x[[i]], ...)
157 mtext(attr(x, "labels"), side = 2, at = 1:n, las = 1)
160 prop.clades <- function(phy, ..., part = NULL)
164 if (length(obj) == 1 && class(obj[[1]]) != "phylo")
165 obj <- unlist(obj, recursive = FALSE)
166 part <- prop.part(obj, check.labels = TRUE)
168 bp <- .Call("bipartition", phy$edge, length(phy$tip.label),
169 phy$Nnode, PACKAGE = "ape")
170 if (!is.null(attr(part, "labels")))
171 for (i in 1:length(part))
172 part[[i]] <- sort(attr(part, "labels")[part[[i]]])
173 bp <- lapply(bp, function(xx) sort(phy$tip.label[xx]))
174 n <- numeric(phy$Nnode)
175 for (i in 1:phy$Nnode) {
176 for (j in 1:length(part)) {
177 if (identical(all.equal(bp[[i]], part[[j]]), TRUE)) {
178 n[i] <- attr(part, "number")[j]
187 boot.phylo <- function(phy, x, FUN, B = 100, block = 1, trees = FALSE)
189 if (is.list(x) && !is.data.frame(x)) {
190 if (inherits(x, "DNAbin")) x <- as.matrix(x)
196 x <- matrix(x, n, nL/n, byrow = TRUE)
200 boot.tree <- vector("list", B)
203 y <- seq(block, ncol(x), block)
204 boot.i <- sample(y, replace = TRUE)
205 boot.samp <- numeric(ncol(x))
206 boot.samp[y] <- boot.i
207 for (j in 1:(block - 1))
208 boot.samp[y - j] <- boot.i - j
209 } else boot.samp <- sample(ncol(x), replace = TRUE)
210 boot.tree[[i]] <- FUN(x[, boot.samp])
212 for (i in 1:B) storage.mode(boot.tree[[i]]$Nnode) <- "integer"
213 storage.mode(phy$Nnode) <- "integer"
214 ans <- attr(.Call("prop_part", c(list(phy), boot.tree),
215 B + 1, FALSE, PACKAGE = "ape"), "number") - 1
217 class(boot.tree) <- "multiPhylo"
218 ans <- list(BP = ans, trees = boot.tree)
223 consensus <- function(..., p = 1, check.labels = TRUE)
225 foo <- function(ic, node) {
227 ## node: node number in the final tree
230 for (j in (ic + 1):m) {
231 wh <- match(pp[[j]], pool)
232 if (!any(is.na(wh))) {
233 edge[pos, 1] <<- node
235 edge[pos, 2] <<- nextnode <<- nextnode + 1L
243 ind <- pos:(pos + size - 1)
244 edge[ind, 1] <<- node
245 edge[ind, 2] <<- pool
250 if (length(obj) == 1) {
251 ## better than unlist(obj, recursive = FALSE)
252 ## because "[[" keeps the class of 'obj':
254 if (class(obj) == "phylo") return(obj)
256 if (!is.null(attr(obj, "TipLabel")))
257 labels <- attr(obj, "TipLabel")
259 labels <- obj[[1]]$tip.label
260 if (check.labels) obj <- .compressTipLabel(obj)
263 ## Get all observed partitions and their frequencies:
264 pp <- prop.part(obj, check.labels = FALSE)
265 ## Drop the partitions whose frequency is less than 'p':
266 pp <- pp[attr(pp, "number") >= p * ntree]
267 ## Get the order of the remaining partitions by decreasing size:
268 ind <- sort(unlist(lapply(pp, length)), decreasing = TRUE,
269 index.return = TRUE)$ix
273 edge <- matrix(0L, n + m - 1, 2)
282 structure(list(edge = edge, tip.label = labels,
283 Nnode = m), class = "phylo")