Add MCITree(): Maximum Clade Information tree selection (#191)

Selects the tree from a posterior sample with the highest total split
information content — analogous to Maximum Clade Credibility but using
phylogenetic or clustering information content.

Supports three modes: phylogenetic (default), clustering, and
credibility (MCC-equivalent). Uses only existing TreeTools/TreeDist APIs
(SplitFrequency, SplitwiseInfo, ClusteringInfo) — no dependency on the
unreleased TreeTools .CountSplits branch.

Author: ms609

NAMESPACE

@@ -84,6 +84,7 @@ export(LnTreesConsistentWithTwoSplits)
 export(Log2TreesConsistentWithTwoSplits)
 export(MASTInfo)
 export(MASTSize)
+export(MCITree)
 export(MSTSegments)
 export(MapTrees)
 export(MappingQuality)
@@ -175,6 +176,7 @@ importFrom(TreeTools,Preorder)
 importFrom(TreeTools,RenumberTips)
 importFrom(TreeTools,RootOnNode)
 importFrom(TreeTools,RootTree)
+importFrom(TreeTools,SplitFrequency)
 importFrom(TreeTools,SplitInformation)
 importFrom(TreeTools,SplitsInBinaryTree)
 importFrom(TreeTools,TipLabels)

NEWS.md

@@ -1,5 +1,11 @@
 # TreeDist 2.13.0 (2026-03-17)
 
+## New features
+
+- `MCITree()` selects the tree from a posterior sample with the highest
+  total split information content — a Maximum Clade Information analogue
+  of the Maximum Clade Credibility tree.
+
 ## Performance
 
 Pairwise distance computation has been substantially optimized.

R/tree_information.R

@@ -379,3 +379,77 @@ ConsensusInfo <- function(trees, info = "phylogenetic", p = 0.5,
   }
   consensus_info(trees, mode == 1L, p = safeP)
 }
+
+#' Maximum Clade Information Tree
+#'
+#' Analogous to the Maximum Clade Credibility tree:
+#' select the tree from a posterior distribution whose clades have the
+#' highest information content.
+#' Generate the MCC tree by specifying `info = "credibility"`.
+#'
+#' @inheritParams ConsensusInfo
+#' @return `MCITree()` returns the tree with the highest information
+#' content, selected from `trees`.
+#' @examples
+#' library("TreeTools", quietly = TRUE)
+#' trees <- as.phylo(24:40, 16)
+#'
+#' # Maximum Clade Information tree
+#' mci <- MCITree(trees)
+#' SplitwiseInfo(mci)
+#' plot(mci)
+#' p <- SplitFrequency(mci, trees) / length(trees)
+#' LabelSplits(mci, round(p * 100), "%", bg = SupportColor(p))
+#'
+#' \donttest{
+#' # Compare with Maximum Clade Credibility tree
+#' mcc <- MCITree(trees, "credibility")
+#' plot(mcc)
+#' p <- SplitFrequency(mcc, trees) / length(trees)
+#' LabelSplits(mcc, round(p * 100), "%", bg = SupportColor(p))
+#' SplitwiseInfo(mcc)
+#' }
+#'
+#' @template MRS
+#' @importFrom TreeTools as.Splits SplitFrequency
+#' @export
+MCITree <- function(trees, info = "phylogenetic", check.tips = TRUE) {
+  mode <- pmatch(tolower(info),
+                 c("phylogenetic", "clustering", "credibility",
+                   "spic", "scic"))
+  if (is.na(mode)) {
+    stop("`info` must be \"phylogenetic\", \"clustering\" or \"credibility\"")
+  }
+  if (inherits(trees, "phylo")) {
+    return(trees)
+  }
+  if (length(trees) == 1L) {
+    return(trees[[1]])
+  }
+  if (check.tips) {
+    trees <- RenumberTips(trees, trees[[1]])
+  }
+
+  nTree <- length(trees)
+  treeSplits <- as.Splits(trees)
+  InfoFn <- switch(mode,
+    function(sp, p) SplitwiseInfo(sp, p, sum = TRUE),
+    function(sp, p) ClusteringInfo(sp, p, sum = TRUE),
+    function(sp, p) sum(p),
+    function(sp, p) SplitwiseInfo(sp, p, sum = TRUE),
+    function(sp, p) ClusteringInfo(sp, p, sum = TRUE)
+  )
+
+  treeInfo <- vapply(treeSplits, function(sp) {
+    p <- SplitFrequency(sp, trees) / nTree
+    InfoFn(sp, p)
+  }, double(1))
+
+  chosen <- which.max(treeInfo)
+  nBest <- sum(treeInfo == max(treeInfo))
+  if (nBest > 1L) {
+    message(nBest, " trees tied for best score; returning tree ", chosen)
+  }
+
+  trees[[chosen]]
+}

inst/WORDLIST

@@ -23,9 +23,11 @@ JV
 Jaccard
 Kaski
 Kawa
+KendallColijn
 LAPJV
 LSAP
 Leeuwen
+MCC
 MDS
 MKL
 MacKay
@@ -37,29 +39,28 @@ Meila
 Mol
 Munkres
 MutualClusteringInfo
-Nura
 NNI
 NNIDist
+Nura
 NyeTreeSimilarity
 OEIS
 ORCID
 OpenMP
+POPCNT
 Perotti
 PhysRevE
 PlotTools
-POPCNT
 R's
-Rcpp
 RStudio
 Rcpp
 RdMacros
 RdPack
 Regraft
+SPI
+SPR
 Sammon
 Sammon's
 Soneson's
-SPR
-SPI
 Stamatakis
 TBR
 TBRDist
@@ -78,6 +79,7 @@ Yi
 ac
 aways
 bipartitions
+branchless
 btaa
 bti
 codecov
@@ -98,8 +100,8 @@ hypervolumes
 ingroup
 interdecile
 jonker
-magiclogic
 macOS
+magiclogic
 mergesort
 molbev
 msw
@@ -116,6 +118,8 @@ roy
 scic
 sensu
 shinyjs
+speedup
+speedups
 splitwise
 syab
 sysbio

man/MCITree.Rd

@@ -0,0 +1,52 @@
+test_that("MCITree returns one of the input trees", {
+  trees <- as.phylo(24:40, 16)
+  mci <- MCITree(trees)
+  expect_true(any(vapply(trees, all.equal, TRUE, mci) == "TRUE"))
+})
+
+test_that("MCITree modes give valid results", {
+  trees <- as.phylo(24:40, 16)
+  for (mode in c("phylogenetic", "clustering", "credibility")) {
+    result <- MCITree(trees, mode)
+    expect_s3_class(result, "phylo")
+  }
+})
+
+test_that("MCITree aliases match base modes", {
+  trees <- as.phylo(24:40, 16)
+  expect_equal(MCITree(trees, "spic"), MCITree(trees, "phylogenetic"))
+  expect_equal(MCITree(trees, "scic"), MCITree(trees, "clustering"))
+})
+
+test_that("MCITree handles edge cases", {
+  tree <- PectinateTree(8)
+  expect_identical(MCITree(tree), tree)
+
+  trees <- c(PectinateTree(8), BalancedTree(8))
+  expect_identical(MCITree(trees[1]), trees[[1]])
+})
+
+test_that("MCITree rejects invalid info", {
+  trees <- as.phylo(24:40, 16)
+  expect_error(MCITree(trees, "invalid"), "must be")
+})
+
+test_that("MCITree credibility selects highest-support tree", {
+  # 10 copies of tree A + 1 copy of tree B:
+  # tree A's splits have frequency 10/11; tree B's have ~1/11.
+  # The credibility tree should be one of the A copies.
+  treeA <- PectinateTree(8)
+  treeB <- BalancedTree(8)
+  trees <- c(rep(list(treeA), 10), list(treeB))
+  class(trees) <- "multiPhylo"
+  mcc <- MCITree(trees, "credibility")
+  expect_equal(mcc, treeA)
+})
+
+test_that("MCITree messages on ties", {
+  # Identical trees should tie
+  tree <- PectinateTree(8)
+  trees <- rep(list(tree), 5)
+  class(trees) <- "multiPhylo"
+  expect_message(MCITree(trees), "tied")
+})