How do I interpret Maaslin3 summary plot

Downstream analysis and statistics
1

Hello,

This might be a super dumb question, but I couldn’t really find any “dumbed down” explanations anywhere online.

I was wondering how do I interpret the summary plot of Maaslin3. For example, the plot used in the nature article, let us only focus on the top taxa - clostridium sp.AT4.

  1. Would I be correct in saying that it is significantly differentially abundant/prevalent in CD but not in UC vs healthy control, after accounting for abx and read depth?
  2. How do I interpret the heatmap? For prevalence, it is lower in non-IBD and higher in CD and UC?
  3. I see that it is ** for read_depth as well as a positive beta coefficient. Does that mean this taxa is diff. abundant as a result of read_depth?

Sorry for these dumb questions. I’m a wet lab scientist trying to start and navigate the world of bioinformatics.

image
image1771×1929 432 KB

2

Hi,

Not a dumb question at all!

  1. Based on the first row, the taxon has significantly increased abundance in CD dysbiosis (but not significantly different prevalence since the triangle is not dark green). This is indeed after controlling for read depth and antibiotic usage. The taxon is not significantly different in either abundance or prevalence for UC dysbiosis.
  2. The heatmap works the same way as the coefficient plots on the left, but it’s just more compact and used for variables you care less about or that weren’t as significant as the ones in the coefficient plot. In this case, based on the red and blue, prevalence is higher in UC and CD (though not significantly so) than in non-IBD (which is the control). Since non-IBD is the control against which everything is compared, there’s no grid cell for IBD because all the grid cells represent a difference based on some variable.
  3. Yes - what you see here is pretty standard: most taxa will have positive significant prevalence associations with read depth since it becomes easier to detect a taxon with deeper sequencing. This is why we recommend always including read depth in the MaAsLin formula.

Will