mgcluster version issues

Commands in mothur
1

I am trying to use mgcluster with my 454 metagenomic datasets and have two issues relating to different versions of mothur:

  1. With mothur 1.18.1 (Mac 64 bit), I get the following error when running mgcluster:
    mothur > mgcluster(blast=AMD.blast)

Reading names… 99419 names read.
********************###########
Reading blast: ||ABError: Sequence ‘003459_AMD_XYG.b2.gene1’ was not found in the names file, please correct

If I provide a names file, I can get around this problem, but this seems like a bug because a names file shouldn’t be necessary, right? When I use mothur 1.15.0 on the same machine, this error is not raised.

  1. If I provide a names file, mglcuster runs successfully on mothur 1.18.1, but the resulting .list file doesn’t compute all of the distances. For one file, it stopped at 0.24. A smaller file stopped at 0.29. If I use mothur 1.15.0, distances up to 0.70 are computed.

I am fine with using 1.15.0 for now unless you know of any issues with using that version of mgcluster.

Billy Brazelton

2

The names file should not be needed. We will have to look into that. The change in distances outputted is most likely due to a change in the default clustering algorithm for version 1.18.1. The default was furthest and now it’s average.

3

Yes, you’re right. I specified method=average with version 1.15.0, and the distances stopped at 0.24. And I just realized for the first time that the .fn and .an extensions on the filenames refer to average neighbor and furthest neighbor! I always wondered about that…

4

I am having a hard time reproducing the sequence name issue you were having. Would you mind sending your blast file to mothur.bugs@gmail.com?

5

I could email you the file, but I get the error even when using the AMD.blast file I downloaded from http://www.bio.umass.edu/micro/schloss/metaG_tools/.