indicator species analysis

[peziza]

is there an indicator species analysis, or something similar, available in Qiime? thanks!

[Jai Ram Rideout]

Hello,

Currently, QIIME does not have indicator species analysis. We are
considering adding indicator species analysis (ISA) and/or
discriminant function analysis (DFA) to QIIME in the future.

What type of questions/tests are you interested in performing using
indicator species analysis? QIIME contains a script called
otu_category_significance.py, which can perform ANOVA, G-test of
independence, and Pearson correlation given an OTU table and a mapping
file. This script may or may not be useful to you- have a look at the
script and please let us know if you have any questions about it. I
think the G-test of independence may be of the most interest to you.

Outside of QIIME, R has indicator species analysis functions in the
labdsv package. The relevant functions are 'indval' and 'duleg'.

R's klaR package has a function 'stepclass' that will allow you to
perform a stepwise selection of OTUs that best predict the
grouping/classification that you provide in the mapping file (you can
specify 'lda' for linear DFA, and the function also accepts other
classification functions).

Please let me know if you have any additional questions.

Thanks,
Jai

[Jai Ram Rideout]

Hi again,

QIIME also has a script called supervised_learning.py that may be of
interest to you. There is also an accompanying tutorial:
http://qiime.org/tutorials/running_supervised_learning.html

In particular, one of the output files called
'feature_importance_scores.txt' will contain the discriminative power
of each OTU in determining what class/group a sample will fall into.
This seems similar to the type of information given by discriminant
function analysis (DFA).

-Jai

[peziza]

Thank you! This was all very helpful!!

On Wednesday, July 11, 2012 10:07:50 AM UTC-7, Jai Rideout wrote:

Hi again,

QIIME also has a script called supervised_learning.py that may be of
interest to you. There is also an accompanying tutorial:
http://qiime.org/tutorials/running_supervised_learning.html

In particular, one of the output files called
'feature_importance_scores.txt' will contain the discriminative power
of each OTU in determining what class/group a sample will fall into.
This seems similar to the type of information given by discriminant
function analysis (DFA).

-Jai

On Wed, Jul 11, 2012 at 9:57 AM, Jai Ram Rideout 

> Hello,
>
> Currently, QIIME does not have indicator species analysis. We are
> considering adding indicator species analysis (ISA) and/or
> discriminant function analysis (DFA) to QIIME in the future.
>
> What type of questions/tests are you interested in performing using
> indicator species analysis? QIIME contains a script called
> otu_category_significance.py, which can perform ANOVA, G-test of
> independence, and Pearson correlation given an OTU table and a mapping
> file. This script may or may not be useful to you- have a look at the
> script and please let us know if you have any questions about it. I
> think the G-test of independence may be of the most interest to you.
>
> Outside of QIIME, R has indicator species analysis functions in the
> labdsv package. The relevant functions are 'indval' and 'duleg'.
>
> R's klaR package has a function 'stepclass' that will allow you to
> perform a stepwise selection of OTUs that best predict the
> grouping/classification that you provide in the mapping file (you can
> specify 'lda' for linear DFA, and the function also accepts other
> classification functions).
>
> Please let me know if you have any additional questions.
>
> Thanks,
> Jai
>
> On Wed, Jul 11, 2012 at 1:17 AM, peziza