Distance Matrix For Metagenomes

Question

[blanca]

Hi,

I would like to compare metagenomes and I have though about making a distance matrix. What I am not quite sure is if it would be a goog idea to create it based on a blast (old blastall function). In that case, it shouldn't be any differences between considering one or another as the database, right? I have just tried and I don't get the same results if I blast A to B or B to A. Does anyone have an idea why could it be? Thank you for your help in advance.

[5heikki]

Consider metagenome I (a,b,c) and metagenome II (x,y).

best hits (I vs II):
a > x
b > x
c > y

best hits (II vs I)
x > a
y > c

Anyway, I don't think this approach makes any sense. How exactly are you going to transfer all vs all into a distance matrix where just one number describes the distance? Cumulative e-values? Cumulative bitscores? Makes no sense at all. Something like weighted unifrac makes a lot more sense, but I'm not sure how that is going to work with wgs. Then there's clustering of taxa abundance heatmaps..

Answer 1

[Frédéric Mahé]

Hi,

You could try compareads. That tool gives you similarity values between metagenomic data sets, that you can use to produce visualizations (heatmaps for example).

Answer 2

[blanca]

Thank you, it seems I need to carefully rethink my aproach. I would try unifrac and compareads!