different outputs of "metastats"

Commands in mothur
1

Hi!
I have been used mothur for a long time and got a lot of benefits from it. I’m grateful to the software developmemt team. Here are several questions on metastats.

I applied metastats command in mothur to detect differentially abundant otus from Healthy and Unhealthy populations in all 30 oral samples (pyrosequencing data).

As the tutorial I took the .shared and .design file as input,the command is as following:

metastats(shared=pg1_r_0.97.shared, design=pg1_r.design)

The output file seem to be that as requested: there were 579 otus were detected to be significantly(p<0.05) different between H and U populations.
For the purpose of getting q value of each otu, I fomatted the file of pg1_r_0.97.shared for the original software on website (http://metastats.cbcb.umd.edu/detection.html)and submitted it to re-run this data.

However, some questions came out when I compared the data analysis results from different versions of “metastats”:

  1. the total number of the otus detected shifted to 459.
  2. most of p values of otus detected were different from those in the output of mothur version

Metastats in mothur
OTU mean(group1) variance(group1) stderr(group1) mean_of_counts(group1) mean(group2) variance(group2) stderr(group2) mean_of_counts(group1) p-value
6 3.083369 6.606199 0.686929 4.714286 0.461625 1.765457 0.332176 0.9375 0.001998
7 0.063472 0.028229 0.044904 0.142857 0.352779 0.16253 0.100788 0.75 0.015984
11 0 0 0 0 0.205585 0.676243 0.205585 0.375 0.000999
…

Metastats on website
Name mean(group1) variance(group1) std.err(group1) mean(group2) variance(group2) std.err(group2) pvalue qvalue
Otu6 0.000111782 9.66E-08 8.03E-05 0.000653306 3.49E-07 0.000152491 0.002997003 0.049778899
Otu7 8.54E-05 8.49E-09 2.38E-05 1.34E-05 1.36E-09 9.54E-06 0.013753493 0.132860173
Otu11 4.98E-05 3.72E-08 4.98E-05 0 0 0 0.031883871 0.216914569
…

Have you found this before? Is there any difference from the mothur version of metastats and the original ?



In addition, why mothur didn't provide the q value as it should be useful if the data is complex(many taxa or otus)?
Thanks again :)
2

The version of metastats in mothur uses the source code provided by White and Pop. The only changes we have made is to the memory handling which doesn’t affect the output. The source code didn’t have correction to get the q-value, but we’re working on adding that. Quickly looking at the p-values they look within the range you’d expect considering there is a randomization procedure.

Hope this helps,
Pat