Unifrac and NMDS

Use this forum to address questions about the use of various commands available in mothur
Post Reply
jitendrakeshri
Posts: 22
Joined: Fri Nov 27, 2015 9:41 am

Unifrac and NMDS

Post by jitendrakeshri » Thu Apr 20, 2017 7:16 am

In core microbiome analysis, only 3 OTUs out of 1232 OTUs were shared among all samples at the relative abundance of 0.01.
We have two groups of samples which are significantly different on the basis unifrac weighted analysis.
But in NMDS plot based on thetayc distance does not show clustering of these groups.

Reviewer has few query:

1. How was the unifrac analysis possible when the communities do not share a significant core microbiome?

2. Why NMDS does not show similar trend as Unifrac analysis?

Any suggestion will be most welcome.

campenr
Posts: 61
Joined: Wed Mar 02, 2016 2:00 pm

Re: Unifrac and NMDS

Post by campenr » Thu Apr 20, 2017 9:34 am

So first point is that distance matrix calculation and ordination are two different things. An NMDS ordination requires a distance matrix as an input, but it is only a visualization technique.

Statistical testing of grouping of samples (you haven't said what method you are using to test this, AMOVA? PERMANOVA?) occurs on the distance matrix, not the ordination itself.

Therefore, in your case you have calculated significant groupings using one distance matrix, but are trying to ordinate using another. I recommend trying to produce an NMDS ordination using the UniFrac distances instead of the Thetayc distances and see what you get (should see grouping of samples if your statisitcal test is indicating significant grouping).

jitendrakeshri
Posts: 22
Joined: Fri Nov 27, 2015 9:41 am

Re: Unifrac and NMDS

Post by jitendrakeshri » Fri Apr 21, 2017 3:41 pm

Thanks Campenr for quick and informative reply. I will try NMDS based on Unifrac metric and see the difference.

Regards,

jitendrakeshri
Posts: 22
Joined: Fri Nov 27, 2015 9:41 am

Re: Unifrac and NMDS

Post by jitendrakeshri » Thu Apr 27, 2017 3:19 am

I have tried NMDS based on unifrac distances, but could not get the desired groupings. Please suggest.

The weighted unifrac analysis between lakes is as:

Tree# Groups WScore WSig
1 High FIC lakes - Low FIC lakes 0.679546 <0.0010

Regards,

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest