I'm publishing my results (hooray!) in a journal that requires individual datapoints to be displayed overlayed on bar graphs. I'm including the results of an ANCOM-BC analysis and I'm wondering if there's a way to get the individual log fold-change differentials calculated for each replicate.
I'm able to extract the raw csv files from the differentials qza file, but it appears that only the average LFC is shown for each group (see attached)
As a side note, I'm seeing far more ASVs in the LFC csv than appear in the bar plot visualization that's made. I suspect that's because I used the --p-conserve "True" parameter, and a lot of the ASVs are getting filtered out due to low significance, but I can't figure out what the criteria being used to filter them is.
The log fold change in an ANCOM-BC barplot represents a point estimate, and not a mean or other aggregation of data. (It's closer to a forest plot, than a boxplot). It's not really possible to show the individual points that go into the estimate, becuause the estimate may be adjusted (for example if your model is age + disease and disease has a large impact, you might see a bimodal distribution within the data, which does not reflect the estimate.
There are a couple of options to consider.
Re plot the full data as a volcano plot using your favorite software. (Excel works well, as do R, python, etc). This is a common plot in bioinformatics and makes it clear you have a fit estimate, rather than a specific value.
Make your captions, methods, and text clear that the barplot is a fit value rather than a mean or forest plot and it's an adjusted fit value.
In either case, provide a plot of the appropriately transformed data associated with yout outcome. (For example, a boxplot, raindrop plot, etc) as appropriate for your sample size and field of the statistically significant taxa.
Put the full file in the supplement so other people can find non-significant taxa.
IIRC, only the significant taxa are shown in the barplot. Your LFC file contains 360 taxa which is way too many to visualzie that way. (Again, a volcano plot can let you see all of them).