he single bar for “All samples” shows that around half of all taxa are shared (3584), but among the remaining half many more taxa are unique to contigs (2753) than to reads (771).
As previously discussed, it could be interesting to divide these taxa fractions into bacteria, fungi, metazoa, viruses, etc. Maybe there is a trend...
The PCoA plot shows that community composition is driven mainly by between-sample differences rather than by the processing method.
I find this quite interesting and I think maybe we should further try to understand why in some cases the similarity between contig-based & read-based tax annotation in samples is greater (points far away from each other of same sample) and in some cases it is lower (points close to each other in PCoA of the same sample). I there anyway to explain this?
Maybe it could be also interesting to color points according to season and use instead of red and blue forms like circle and square of the same color to understand if differences between read-based and contig-based approaches are influenced by seasonality.