is in qiime possible to perform a 2ways anova testing treatments and sampling time?

[Valentina Imparato]

Hi there! I have another doubts how to proceed with my analysis to test significance between the treatments I have and depending on time of the incubation. I have split the otu tables according to treatments and time and tested individually by group_significance.py. but I am wondering if there is any script that allows me to test the 2 factors together.  let's say I would like to compute the significance on the abundance depending on the treatment but I have independent sampling along time scale. so I would like also to know if the incubation time is affecting the variation. there is a way how to test time series samples in Qiime? in the table a little portion of my samples for you to see.  let's say that 37-42 are triplicates of 1 treatment relative to 2 time points, 43-48 the control relative to 2 time points as well. the complete set has 4 treatments and 3 time points. I have the 1.8 version of qiime. Thanks you folks for help. all the best, Vale EXTRACT OF TABLE from summarize_taxa (trimming the rare otu) Taxon VAL037 VAL038 VAL039 VAL040 VAL041 VAL042 VAL043 VAL044 VAL045 VAL046 VAL047 VAL048 Bacteria;Acidobacteria;Acidobacteria_Gp4;Gp4;Other;Other 462 640 639 504 900 753 1046 632 587 323 547 467 Bacteria;Acidobacteria;Acidobacteria_Gp6;Gp6;Other;Other 943 1058 1482 1452 1804 1399 1371 1019 806 600 952 966 Bacteria;Acidobacteria;Acidobacteria_Gp6;Other;Other;Other 241 249 359 364 471 318 345 199 231 125 221 254 Bacteria;Actinobacteria;Other;Other;Other;Other 193 282 255 411 431 275 166 212 97 164 243 189 Bacteria;Bacteroidetes;Bacteroidetes_incertae_sedis;Other;Other;Other 157 163 154 96 147 160 111 114 54 85 67 43

[Jenya Kopylov]

Hi Valentina,

After discussing with Tony, you may need to use R for this. This link may be helpful. You can convert the OTU table into a tab-delimited format using Excel.

Jenya

[Will Van Treuren]

Hi Valentina, 

Too add to Jenya's post, you can do a version of this test in QIIME by manipulating your metadata.  What you need to do is create a metadata column in your mapping file that is the concatenation of the two fields you are interested in (so treatment and time based on your previous post). You can create a new column in your mapping file, and populate is with something like TREATMENT_TIME where TREATMENT is the value under the treatment metadata column, and TIME is the value under the sampling time column (an excel formula like =concatenate(A2, "_", B2) will do the trick). 

Once you have this column, you can run group_significance.py against it, and it will tell you if any treatments are different based on this grouping. It is not as good as a normal two way ANOVA because it won't give you the p-values for the individual factors or interactions, but it might be a start without having to go to R. 

Hope this helps,

Will 

[Barvaz Sini]

Hi,

Can you also try to a bit more exactly what you are looking for? Are you looking for a difference between the 2 treatments in any single time point? or in the final outcome? or a trend? what type of effect are you looking for?

Thanks,

Amnon