Tuesday, June 28, 2016

Fix for bug in collapseTree

Not long after submitting a new version of phytools (phytools 0.5-38) to CRAN, which has been accepted, I discovered a weird bug in the interactive function collapseTree. collapseTree is interactive, so it is kind of hard to demo; however, a video of it's use can be seen here.

The bug is due to the plotrix function draw.circle either no longer being vectorized or not being correctly vectorized. I essentially fixed this by substituting:




This fix is on GitHub and can be installed in the typical way.

Friday, June 24, 2016

New version of phytools (0.5-38) on CRAN

There is a new version of phytools (0.5-38) on CRAN in advance of the workshop I will teach this coming week in San Juan, Puerto Rico with Mike Alfaro, Ricardo Betancur, and Luke Harmon.

Here are some of the updates for this version:

  1. A simple function (matchLabels) to match labels between trees.

  2. A new function (consensus.edges) containing various methods to compute the consensus edge lengths from a set of trees.

  3. Fixes to the functions dotTree and fitPagel (here). (Also here: 1, 2.)

  4. An update to evol.vcv to fit a one matrix only model. (I also subsequently pushed an update to allow it work for only one trait too.)

  5. New S3 method updates for object classes "simmap" and "multiSimmap" and new function mapped.states (here).

  6. More important fixes to dotTree for trees containing few taxa.

  7. Some internal (but important) ways in which the way phytools plots left-facing trees. (Also here.)

  8. New function, averageTree for computing the 'average' phylogeny in a set, by some criteria (1, 2, 3, 4).

  9. Separate user control of left- and right-facing tree font sizes in plot.cophylo (here).

  10. Update to cophylo to permit links in assoc for taxa not found in the trees (here).

  11. Simple as.phylo method for objects of class "simmap" and "multiSimmap" (here).

  12. New options for setting the root state in sim.history.

  13. A version of bind.tree for objects of class "simmap" (also here).

  14. Important updates to fitPagel to permit different dependency relationships between traits and different models of evolution.

  15. New options for user control in dotTree (here).

  16. A new S3 plot method for objects of class "fitPagel" (1, 2).

  17. S3 plotting method for discrete character models fit using fitMk or fitDiscrete in the geiger package (1, 2, 3).

  18. Important updates to the plot.phylo.to.map method (here).

  19. Update to function (multiC) for computing multiple phylogenetic VCV matrices depending on the state of a mapped trait (here).

  20. Important bug fix for the function contMap.

  21. Update to roundPhylogram to permit different line edges and types.

  22. Finally, a new interactive version of phylo.toBackbone to permit a backbone tree with subtrees plotted as triangles to be created interactively, rather than via a static translation table (here).

These updates are already on CRAN - but it will probably take a couple of days for binaries to be built and percolate through the mirror repositories. In the meantime, as always, phytools can be installed from GitHub. Enjoy!

Interactive version of phylo.toBackbone

I just created a new interactive version of the phytools function phylo.toBackbone, which creates a backbone phylogeny in which some clades are represented as triangles. This tree can then we plotted with an S3 plot method for the object class "backbonePhylo". Most of the code changes can be seen here. The updates are kind of hacky - principally because the original version of this appears to have been coded by me in 2013.

In previous versions of this function (e.g., 1, 2) the user had to supply the backbone tree as input along with a translation table. This is still possible, or course; however the current version allows the user to collapse clades into triangles by clicking on the fully plotted tree. This should be a nice feature for some users.

Obviously, interactive mode is kind of difficult to demonstrate - consequently I have embedded a screenshot video of a demo at the bottom. The following is just an illustration of the product. Remember for publication to save to PDF (or another line art format) to avoid aliasing, particularly on the edges of the triangles.

## [1] '0.5.38'
## Backbone phylogenetic tree with 15 subtrees and 14 resolved internal nodes.
## Labels: t40, t10, t27, t13, t39, ...
## Diversities: 1, 1, 1, 1, 1, ...
plot(obj) ## defaults

plot of chunk unnamed-chunk-1

cols ## color some of the clades
##  Clade 1  Clade 2  Clade 4 
##  "black"    "red" "green3"
par(cex=0.8) ## change font size
par(font=3) ## change font type

plot of chunk unnamed-chunk-1

Here is the interactive demo:

Saturday, June 11, 2016

Update to roundPhylogram to permit different lwd and lty for different edges

I just pushed a small update to roundPhylogram that permits different line types (lty) and line widths (lwd) to be specified for different branches of the tree. This will be done whenever either argument is supplied as a vector instead of as a single scalar value or (in the case of lty) as a character string.

I did this so that I could plot taxa with ambiguous phylogenetic placement (say) using a dotted line. Note that it would also be straightforward to do this in functions using plotSimmap or the internal function phylogram, but this does not yet exist. It may exist in the generic S3 plot method for objects of class "phylo". I don't know, but I wouldn't be surprised.

Here is what I mean.

packageVersion("phytools") ## latest GitHub version
## [1] '0.5.37'
## read trees


mtext("    A)",side=3,adj=0,line=1.2,cex=1.2)
mtext("    B)",side=3,adj=0,line=1.2,cex=1.2)

plot of chunk unnamed-chunk-1

Note that the consensus of these two trees (similar as they are in so many respects, except with respect to the placement of "D"), is a star tree:


plot of chunk unnamed-chunk-2

Tuesday, June 7, 2016

Fix to contMap for short edges descended from the root node

A phytools user, Bruno Vilela, recently reported the bug that the phytools function contMap can fail for very short internal edges. This tends to be the case when an edge has a length that is less than 1/res × the tree height, in which res is an argument of contMap specifying the graininess of the color gradient to be plotted along edges.

[Actually, I believe I already fixed this bug except for edges descended from the root node. This is because for those edges the function will attempt to create a list of mappings along edges, but create a vector instead. Silly!]

I have previously suggested collapsing very short internal edges or increasing res; however Bruno did one better by providing an incredibly simple solution, which was to create a list of mappings along edges before iterating across the edges of the tree to compute the mappings. Perhaps obvious in retrospect, please keep in mind that I wrote the first version of this function in 2012 when I was quite an R programming novice. (I'm not sure what I'd call myself now. Slightly more than an a novice? I guess in an era when Donald Trump is running for present I'm a world-class R programming expert!)

Here's a quick demo of the problem, along with a demonstration that Bruno's fix does indeed restore function:

## Phylogenetic tree with 26 tips and 25 internal nodes.
## Tip labels:
##  t5, t26, t18, t17, t19, t22, ...
## Rooted; includes branch lengths.
##          t5         t26         t18         t17         t19         t22 
##  0.20063042  0.90298749  2.28555569  2.11331867  1.38012423  0.01110355 
##         t15         t12          t6          t3          t2         t24 
##  0.31029532 -0.31270146  0.36006396  0.86328160 -0.16281977 -1.88665386 
##          t4          t9         t25         t11          t8         t21 
## -2.40507772 -2.69645998 -1.57485299  1.30619095  1.96062919 -0.91760839 
##          t1         t14          t7         t16         t20         t23 
## -1.45220994 -2.58959187 -1.44919866  0.93419977  0.19530914  0.06853757 
##         t13         t10 
##  1.79467390  0.29857041
plotTree(tree) ## short node from the root

plot of chunk unnamed-chunk-1

## Error in tree$maps[[i]] <- XX[, 2] - XX[, 1]: more elements supplied than there are to replace
## increase res

plot of chunk unnamed-chunk-1

Now let's load the new version from source off GitHub:


plot of chunk unnamed-chunk-2

This fix also allows us to set the value of res to any arbitrarily low value. Although the resultant plot is grainier, this might nonetheless be useful for users working with very large trees where computation is an issue.


plot of chunk unnamed-chunk-3


plot of chunk unnamed-chunk-3

Thanks Bruno!