Unweighted UniFrac Randomization Problems

I’ve been trying out weighted and unweighted UniFrac on the AbRecovery dataset and have been getting some strange results with the latest release (1.13.0). I loaded the abrecovery.paup.nj tree and ran the unifrac.unweighted() command. Using the default settings, mothur came up with a UWScore that was pretty close (0.703848) to the value given in the tutorial (0.6818) and a significance score of zero. I ran it again setting random=true to see if the significance matched. After cranking through the 1,000 iterations, the output showed up as:

Tree# Groups UWScore UWSig
1 A-B 0.715765 1
1 A-C 0.726746 1
1 B-C 0.75528 1

Opening the abrecovery.paup.nj.unweighted file shows no iterations, just the same three lines shown in the output above. I’ve run unweighted.unifrac() on a number of my own trees with the same results: normal looking score but always a significance score of 1. Any ideas?

Conversely, unifrac.weighted seems to be working fine.

Thanks for reporting this bug. It seems when we added the paralellization to the unifrac commands in the last release we introduced this bug. The scores for unifrac.unweighted are fine, but the significance is incorrect. The fix will be part of 1.14.0.