Phylogenetic Tree Programs Compatible With MOTHUR

Commands in mothur
1

I’m wondering which phylogenetic tree building programs are compatible with MOTHUR. I would like to be able to perform the parsimony test on a very large pyrosequencing data set (~120,000 sequences split among 5 groups). My distance matrices are upwards of 10 GB. Unfortunately the Clearcut website seems to be down and I cannot obtain that software that has been integrated for use in MOTHUR. I have been looking at FastTree as an alternative but am getting errors about reading the … in my alignment files. Does anybody know if this will cause a problem in downstream in the generation of the tree accuracy?

I would appreciate any suggestions or comments,

Elizabeth

2

I am attaching source code for clearcut 1.0.9. I don’t think James will mind.
I deleted the examples to get under the Mothur forum’s 256kb upload limit.
But honestly, with the huge number of sequences you have, there is an over
9000% chance that tree-based methods will declare your samples to be statistically,
significantly different.

Robin

3

Thanks for the software, I appreciate it. Since the tree building Parsimony Test seems likely to not be the best choice for comparison of libraries, do you think that using LIBSHUFF would be a a better approach? We first want to see if these libraries are significantly different (as we suspect that some are not) before we go into any other in depth analysis.

Elizabeth

4

Although people’s jaws may drop, if you want to do a hypothesis test, the unifrac.weighted/unweighted test might be the way to go.

5

Elizabeth,

I have been using fasttree with no problems for about a month on some fairly large 16S datasets (>100k reads). The warning about the “.” characters can be dismissed as long as your alignments have minimal “.” (mine have 10-20 on either end). I have run the fasttree output through unifrac within mothur on three different sample sets and the sample distance matrices are comparable to those generated using various OTU-based distance matrices (97% similarity clusters, phylotype levels 4-7 with SILVA). I looked into a variety of tree-building algorithms and fasttree seems to work very well, very quickly, and with minimal headaches. Feel free to contact me if you are having issues with fasttree, though I am no expert.

Craig Nelson