search

ggloor / ALDEx_bioc

ALDEx_bioc is the working directory for updating bioconductor
27 stars 13 forks source link

suspicious pattern in results #36

Closed sformel closed 3 years ago

sformel commented 3 years ago

I'm not sure if there is a more appropriate place for this question, if so, could you please point me there? And thank you in advance!

I've gotten some very suspicious looking results from my data, I'm curious if you might be able to give any insight into what would be driving this "handprint" looking pattern?

I'm not sure what the quasi-symmetry and/or each cluster on the plot might represent. I see similar patterns in Fernandes et al. (Figure 4) but they aren't anywhere near as extreme. I suspect they are indications of a poor model fit, but I don't understand the models well enough to be sure how to investigate further.

All the results are from the same experiment, which had 3 replicates for each treatment. All three followed fairly standard Illumina sequencing protocols and DADA2 or Trinity based workflows.

Soil 16S data

16S auto1

Soil ITS data

ITS auto1

Root Transcriptome

auto1

Test

As a test, I've run the Aldex "selex" data and the phyloseq "Global Patterns" data. The latter had similar patterns to my data, but not as extreme.

Selex Data

selex1

Global Patterns Data (soil and sediment data)

globalpatterns1

ggloor commented 3 years ago

Hi. this is a very interesting pattern but not unprecedented

It is caused by differential sparsity in individual samples. What you are seeing is the pattern of unique low count (or 0 count) features in each sample. This is what is giving the individual 'rays' in the effect plot.

sformel commented 3 years ago

Thanks you for the explanation! That actually makes sense given my experimental treatments.

ggloor commented 3 years ago

it's only obvious if you have seen it a thousand times :-)