Simulate sample groups

[Puumanamana]

Hello,

I am trying to use SparseDOSSA2 to simulate a metagenome dataset with multiple sample groups (such as simple case/control). After looking at the tutorial, it seems I want to use the spike-in feature. However, it seems one step is missing from the tutorial (to generate the spike_in_setup dataframe). How would you recommend generating this one?
I also tried simply to do:

n_groups <- 3

metadata <- data.frame(
    Group=rep(1:n_groups, each=500))
)

sim <- SparseDOSSA2(
    template="Stool",
    n_sample=nrow(metadata),
    new_features=FALSE,
    metadata_matrix=as.matrix(metadata),
    verbose=FALSE)

But it doesn't seem the metadata does anything (when I do an NMDS plot of sim$simulated_data, I don't see any grouping). Is it because I need to add something for the spike-in feature_metadata_spike_df argument?

Thank you for your help!

[vstanislas]

Hello,
I think I ran into similar issues as yours and can share the solution that worked for me.
Looking at the code revision history, I found the corresponding part of the code that is removed in the current version:

spike_in_setup <- data.frame(metadata_datum = c(1, 1, 2, 3 ),
                             feature_spiked = c(feature_up, feature_down, feature_up, feature_down),
                             effect_size = c(3, -3, 1 , 1))

Also the feature_metadata_spike_df parameter no longer exists in the current version of SparseDOSSA2(). The df_metadata_spike object can however be read via the spike_metadata parameter.

sim <- SparseDOSSA2(
    template = "Stool",
    n_sample = nrow(mat_metadata), 
    new_features = FALSE,
    spike_metadata = df_metadata_spike,
    metadata_matrix = as.matrix(mat_metadata), 
    verbose = FALSE)

Hope this can help!

Best