Outlier detection

PHLAWD has a procedure that allows for detecting outliers by two methods. To use this feature, after you have run

PHLAWD assemble runfile.phlawd

you can run

PHLAWD outlier runfile.phlawd

This will run a RAxML run on your file generated from the assemble procedure. It assumes that you have something called raxmlHPC-PTHREADS-SSE3 in the path. Customizing this will be an option later.

Options

You can set a number of options related to outlier detection in the runfile.phlawd file.

outliertreefile = filename # file that contains a tree, will skip the raxml step
outliertreerooted # is the tree in the file provided in outliertreefile rooted
taxcutoff = integer # taxcutoff + mean tax distance = tax cutoff, smaller == conservative
blcutoff = integer # blcutoff * mean BL = BL cutoff, smaller == conservative

Output

There are a number of files output from the outlier analyses.

  • genename.taxoutliers – has the gi’s in the first column for the taxonomic outliers
  • genename.taxoutliers.tre – (open in figtree) has branches colored red that were marked as outliers (internal node names are the taxonomic distance of that branch)
  • genename.bloutliers – has the gi’s in the first column for the branch length outliers
  • genename.bloutliers.tre (open in figtree) has branches colored red that were marked as outliers

Note about tree making

The current flags for making the raxml tree just make one tree with GTRCAT. There are a number of ways to make faster trees (with raxml as well). It is recommended that with really large trees, you use another method and then point phlawd in the runfile file to where that tree is so that it doesn’t take forever.

One Comment

  1. Pingback: outlier detection | PHLAWD

Fork me on GitHub