Ratchet: Parsimony Ratchet

MorphyBootstrapR Documentation

Parsimony Ratchet

Description

Ratchet() uses the parsimony ratchet \insertCiteNixon1999TreeSearch to search for a more parsimonious tree using custom optimality criteria.

Usage

MorphyBootstrap(
  edgeList,
  morphyObj,
  EdgeSwapper = NNISwap,
  maxIter,
  maxHits,
  verbosity = 1L,
  stopAtPeak = FALSE,
  stopAtPlateau = 0L,
  ...
)

Ratchet(
  tree,
  dataset,
  InitializeData = PhyDat2Morphy,
  CleanUpData = UnloadMorphy,
  TreeScorer = MorphyLength,
  Bootstrapper = MorphyBootstrap,
  swappers = list(TBRSwap, SPRSwap, NNISwap),
  BootstrapSwapper = if (is.list(swappers)) swappers[[length(swappers)]] else swappers,
  returnAll = FALSE,
  stopAtScore = NULL,
  stopAtPeak = FALSE,
  stopAtPlateau = 0L,
  ratchIter = 100,
  ratchHits = 10,
  searchIter = 4000,
  searchHits = 42,
  bootstrapIter = searchIter,
  bootstrapHits = searchHits,
  verbosity = 1L,
  suboptimal = sqrt(.Machine$double.eps),
  ...
)

MultiRatchet(
  tree,
  dataset,
  ratchHits = 10,
  searchIter = 500,
  searchHits = 20,
  verbosity = 0L,
  swappers = list(RootedNNISwap),
  nSearch = 10,
  stopAtScore = NULL,
  ...
)

RatchetConsensus(
  tree,
  dataset,
  ratchHits = 10,
  searchIter = 500,
  searchHits = 20,
  verbosity = 0L,
  swappers = list(RootedNNISwap),
  nSearch = 10,
  stopAtScore = NULL,
  ...
)

Arguments

edgeList

a list containing the following: - vector of integers corresponding to the parent of each edge in turn - vector of integers corresponding to the child of each edge in turn - (optionally) score of the tree - (optionally, if score provided) number of times this score has been hit

morphyObj

Object of class morphy, perhaps created with PhyDat2Morphy().

EdgeSwapper

a function that rearranges a parent and child vector, and returns a list with modified vectors; for example SPRSwap().

maxIter

Numeric specifying maximum number of iterations to perform in tree search.

maxHits

Numeric specifying maximum number of hits to accomplish in tree search.

verbosity

Numeric specifying level of detail to display in console: larger numbers provide more verbose feedback to the user.

stopAtPeak

Logical specifying whether to terminate search once a subsequent iteration recovers a sub-optimal score. Will be overridden if a passed function has an attribute stopAtPeak set by attr(FunctionName, "stopAtPeak") <- TRUE.

stopAtPlateau

Integer. If > 0, tree search will terminate if the score has not improved after stopAtPlateau iterations. Will be overridden if a passed function has an attribute stopAtPlateau set by attr(FunctionName, "stopAtPlateau") <- TRUE.

...

further arguments to pass to TreeScorer(), e.g. ⁠dataset = ⁠.

tree

A tree of class phylo.

dataset

a dataset in the format required by TreeScorer().

InitializeData

Function that sets up data object to prepare for tree search. The function will be passed the dataset parameter. Its return value will be passed to TreeScorer() and CleanUpData().

CleanUpData

Function to destroy data object on function exit. The function will be passed the value returned by InitializeData().

TreeScorer

function to score a given tree. The function will be passed three parameters, corresponding to the parent and child entries of a tree's edge list, and a dataset.

Bootstrapper

Function to perform bootstrapped rearrangements of tree. First arguments will be an edgeList and a dataset, initialized using InitializeData(). Should return a rearranged edgeList.

swappers

A list of functions to use to conduct edge rearrangement during tree search. Provide functions like NNISwap to shuffle root position, or RootedTBRSwap if the position of the root should be retained. You may wish to use extreme swappers (such as TBR) early in the list, and a more subtle rearranger (such as NNI) later in the list to make incremental tinkerings once an almost-optimal tree has been found.

BootstrapSwapper

Function such as RootedNNISwap to use to rearrange trees within Bootstrapper().

returnAll

Set to TRUE to report all MPTs encountered during the search, perhaps to analyse consensus.

stopAtScore

stop search as soon as this score is hit or beaten.

ratchIter

Stop when this many ratchet iterations have been performed.

ratchHits

Stop when this many ratchet iterations have found the same best score.

searchIter

Integer specifying maximum rearrangements to perform on each bootstrap or ratchet iteration. To override this value for a single swapper function, set e.g. attr(SwapperFunction, "searchIter") <- 99

searchHits

Integer specifying maximum times to hit best score before terminating a tree search within a ratchet iteration. To override this value for a single swapper function, set e.g. attr(SwapperFunction, "searchHits") <- 99

bootstrapIter

Integer specifying maximum rearrangements to perform on each bootstrap iteration (default: searchIter).

bootstrapHits

Integer specifying maximum times to hit best score on each bootstrap iteration (default: searchHits).

suboptimal

retain trees that are suboptimal by this score. Defaults to a small value that will counter rounding errors.

nSearch

Number of Ratchet searches to conduct (for RatchetConsensus())

Details

For usage pointers, see the vignette.

Value

MorphyBootstrap() returns a tree that is optimal under a random sampling of the original characters.

Ratchet() returns a tree modified by parsimony ratchet iterations.

MultiRatchet() returns a list of optimal trees produced by nSearch ratchet searches, from which a consensus tree can be generated using ape::consensus() or TreeTools::ConsensusWithout().

Functions

  • RatchetConsensus(): deprecated alias for MultiRatchet()

Author(s)

Martin R. Smith (martin.smith@durham.ac.uk)

References

\insertAllCited

See Also

  • Adapted from pratchet() in the phangorn package.

Other custom search functions: EdgeListSearch(), Jackknife(), SuccessiveApproximations()

Examples

data("Lobo", package = "TreeTools")
njtree <- TreeTools::NJTree(Lobo.phy)
# Increase value of ratchIter and searchHits to do a proper search
quickResult <- Ratchet(njtree, Lobo.phy, ratchIter = 2, searchHits = 3)

# Plot result (legibly)
oldPar <- par(mar = rep(0, 4), cex = 0.75)
plot(quickResult)
par(oldPar)

ms609/TreeSearch documentation built on March 3, 2024, 5:11 a.m.