Maaslin3 not finding significant features despite clustering in ordination

Downstream analysis and statistics
1

Is it surprising that the same dataset can produce this ordination

image
image791×549 17.6 KB

and yet Maaslin3 would find 0 ASVs differing between these groups?

Balmoral.covariates.csv (345 Bytes)

Balmoral.features.csv (24.2 KB)

test1 ← read.csv(“Balmoral.features.csv”, header = TRUE, row.names = 1)
test2 ← read.csv(“Balmoral.covariates.csv”, header = TRUE, row.names = 1)
test2$Time_Point ← factor(test2$Time_Point)
set.seed(1024)
Balmoral.fit_out ← maaslin3(input_data = test1, input_metadata = test2,
output = ‘repro_output’,
formula = ‘~ Time_Point + Read_depth’,
normalization = ‘TSS’,
transform = ‘LOG’,
augment = TRUE,
standardize = TRUE,
max_significance = 0.1,
median_comparison_abundance = TRUE,
median_comparison_prevalence = FALSE,
max_pngs = 250,
cores = 1)

All associations had errors 
                                or were insignificant.
2

Is the dataset just those 9 samples? I think it’d be really hard to get any significant associations with only 9 samples. Even using dozens of samples typically only yields a few associations for strong effects.

3

This is a subset of a larger dataset, but yes, I suppose you are right about the sample size being the most likely explanation.