search

maracashay / DAWG-Helpline

Need help deciding what step to do or what specific commands to pass when analyzing your amplicon data? Ask AWAY! DAWG is here to help :)
0 stars 1 forks source link

DESeq2 for phylum level? #3

Open ninacamillone opened 4 years ago

ninacamillone commented 4 years ago

For an analysis I've been working on, I had originally used a normal ANOVA to compare relative abundance among samples at higher taxonomic level (only 10 taxa). I also did the same thing for 12 of the most abundant genera.

Someone suggested that I use DESeq2 instead of ANOVA since it's a more standard way of looking at microbiome data. However, now I'm realizing that I'm not sure DESeq2 is appropriate to use at higher taxonomic level, and it seems backwards to select a few key genera using a relative abundance cutoff and then analyze in DESeq2 rather than analyzing first and reporting based on a statistical significance cutoff.

What test/package would you use to look at differential abundance for a few select taxa only (phyla or genera)?

maracashay commented 4 years ago

Can you tell us a little more about your experimental design? I think if your design had random variables, then running a mixed linear effects model would be better than DESeq2.

Most probably what they are thinking is about the distribution of the abundances. DESeq normalizes the abundances and then runs the statistical analysis. ANOVA wouldn't do that, but if your model adheres to the assumptions for linear regression, then I think it is fine.

ninacamillone commented 4 years ago

There are soil microcosms treated with 3 different hydrocarbon substrates individually, in pairs, and all three together. The specific comparisons I'm interested in making here are between the mixed treatments and the individual treatments of those same hydrocarbons (rather than comparing microcosms treated with different substrates or determining the effect of a specific substrate across many treatment combinations etc.) The null hypothesis should be something along the lines of the mixed treatments not being significantly different from (or being intermediate to) the corresponding individual treatments. There are also different concentration levels, but I was only making comparisons within the same concentration in this case. I hope that clarifies my question, and thanks for your input!