Questions about the parameter unit

Downstream analysis and statistics MaAsLin
1

Hi, I’ve recently started using this version and after trying out a few parameters, there are some aspects I don’t quite understand. However, I must say, I’m quite impressed with this version overall. It seems to have many improvements and new features that are quite user-friendly.
1. I have data for three groups (A, B, C), and I am using group C’s data as a reference; however, when I input all three groups (ABC, with C as the reference), the output abundance data results, including the coefficient (coef) values and differential bacteria, are not identical to those generated when I input only two groups (AC, with C as the reference).
2. Additionally, when I input data for the three groups (ABC) with C as the reference, the heatmap on the right side of the final summary plot seems to be unable to clearly show the impact of which group.
3. When I input data for five groups (A, B, C, D, E, F), in the final summary plot, the beta coefficient plot on the left side can only display the results for two of the groups, with the remaining groups being transferred to the prevalence abundance heatmap on the right, where they are mixed together with the results for the other variables (such as timepoints).

Here is the code :thinking::

# used data
taxa_table
metadata

metadata$Treatment <- 
  factor(metadata$Treatment, levels = c('A', 'B', 'C'))
metadata$Tp <- 
  factor(metadata$Timepoint, levels = c('1', '2', '3', '4'))

fit_out <- maaslin3(input_data = taxa_table,
                    input_metadata = metadata,
                    output = 'test_output',
                    formula = '~ Treatment+ Timepoint',
                    normalization = 'TSS',
                    transform = 'LOG',
                    augment = TRUE,
                    standardize = TRUE,
                    max_significance = 0.1,
                    median_comparison_abundance = TRUE,
                    median_comparison_prevalence = FALSE,
                    max_pngs = 100,
                    cores = 1,
                    save_models = TRUE)
2

Hi Darian,

Glad to hear the tool is (mostly) working well!

  1. When you say the two results are not identical, do you mean they’re quite different or they’re close but not exactly the same? There’s a Monte Carlo step in the median comparison step that introduces a small bit of randomness, so that could account for small differences.
  2. What do you mean by “unable to clearly show the impact of which group”? The coefficient plot and heatmap are based on the output coefficients from MaAsLin 3, so if you have multiple categories, each will be tested against the reference.
  3. By default, MaAsLin 3 chooses the two most significant effects to display in the coefficient plot and puts the rest in the heatmap. If there are multiple categories tested against the baseline, this will show up as multiple different coefficient plots or heatmap columns. You can choose which variables are plotted where with heatmap_vars and coef_plot_vars (and can re-plot from outputs).

Will