Skip to content

NEWS.md

Latest commit

 

History

History
102 lines (78 loc) · 4.73 KB

NEWS.md

File metadata and controls

102 lines (78 loc) · 4.73 KB
Raw

PhyloConfigR (development version)

Fixes

  • plot.filterNode() now uses filter.name = "proportion_samples", the name that filterSummary() writes. The old choice, "proportion_sample", matched no rows and stopped with a data.frame() row-count error.

PhyloConfigR 0.3.2

Fixes

  • analysis.concatenationTree() no longer passes -msub on unpartitioned runs (partition.scheme = "none"). -msub only restricts ModelFinder's amino-acid model set, so it has no effect when an explicit model is given, and some IQ-TREE builds reject the flag in that context with ERROR: Unknown sequence type (exit status 2), which made every replicate of a gene jackknife fail at IQ-TREE launch. It is now added only for the ModelFinder schemes ("merge", "file").

  • analysis.concatenationTree() gained a seq.type argument that forces the IQ-TREE data type with -st, and analysis.geneJackknife() now sets it to "DNA" by default. Some IQ-TREE builds fail auto-detection with ERROR: Unknown sequence type on replicate matrices that carry a lot of missing data; forcing the type is deterministic and, if a locus were genuinely not DNA, reports the offending site instead of the opaque error. Set seq.type = NULL to auto-detect.

  • analysis.geneJackknife() now removes each replicate's concatenated matrix even when the replicate fails, not only on success. The matrix is built just before its tree and deleted in a finally block, so only the replicates in flight occupy disk and an aborted run no longer leaves a directory full of matrices behind.

PhyloConfigR 0.3.0

Adds a gene jackknife.

New

  • analysis.geneJackknife() runs a gene jackknife on a folder of locus alignments. Each replicate draws loci at random without replacement to a fixed size (a number of base pairs or a number of genes), concatenates them with concatenateAlignments(), and estimates a tree with analysis.concatenationTree() with UFBoot off, because the resampling is the replication. It builds a majority-rule consensus from the replicate trees, or can be pointed at a single index with replicate.subset so a cluster array runs one replicate per task. Replicate i is seeded from seed + i, so the result does not depend on the order the tasks run and any replicate reproduces on its own. An optional locus.lengths table skips re-reading the alignments on every task.

PhyloConfigR 0.2.1

Fixes to IQ-TREE handling. All four affected functions previously assumed the executable is named iqtree2, which is true of IQ-TREE 2 and not of IQ-TREE 3.

New

  • findIQTREE() locates an IQ-TREE executable and reports its version. Tries iqtree2 then iqtree, and reads the version back from the executable rather than inferring it from the file name. A version that cannot be read is reported as NA rather than being treated as an error, so wrapper scripts and unusual builds still run.

Fixes

  • analysis.concatenationTree() could not be called on its default arguments. partition.scheme and msub.type defaulted to their whole choice vectors, and comparing a length-3 vector with == inside if() is an error in R 4.2 and later. Both now go through match.arg(), and -msub no longer expands to two commands.

  • analysis.concatenationTree() always passed -bb, and IQ-TREE rejects -bb below 1000, so bootstrapping could not be turned off. uf.bootstrap = 0 now omits the flag, matching the convention estimateGeneTrees() already used. This makes the function usable inside resampling procedures that do their own replication, such as a gene jackknife.

  • analysis.concatenationTree() ignored the exit status of IQ-TREE and could report success after a failed run. It now checks the status and that a treefile was produced, and returns the treefile path invisibly.

  • analysis.concatenationTree() gained a model argument for partition.scheme = "none", which previously hard-coded GTR, along with seed, -mem, a working resume, and recursive directory creation.

  • estimateGeneTrees(), concordanceFactors() and concordanceRunner() now resolve the executable with findIQTREE() instead of assuming iqtree2. concordanceFactors() and concordanceRunner() take iqtree.path = NULL by default.

  • NAMESPACE now imports data.table. Listing it under Imports in DESCRIPTION is not enough: data.table's := checks that the calling namespace is data.table-aware, so concatenateAlignments() failed at run time with "[ was called on a data.table in an environment that is not data.table-aware" for every installed copy of the package. Sourcing the R files hid the problem, because the calls then happened in the global environment.

PhyloConfigR 0.2.0

  • Initial packaged release.