Very cool software! I’ve always wanted to compare methods myself but the task seemed to daunting! I believe that it’s important to find a statistical model that works best for a data structure and design similar to one’s own data.
Would it be feasible to simulate data based on the pathway-sample table generated from humann2? Could I simply reformat my pathway table into a QIIME OTU table format and have SparseDOSSA simulate data based on it or is the nature of the pathway data too different from microbiota OTU/ASV type data?
Thanks for the kind words! I don’t see why SparseDOSSA can’t be applied to pathway abundances. What matters is that the input is a) a relative abundance or count matrix of community profiles and b) zero-inflated. Both pathway and taxonomic profiles satisfy these.
If you’d like to be extra cautious, you can always compare the input pathway data with simulated samples from SparseDOSSA (visually by ordination, for example). This should give you an idea of how similar they are and if SparseDOSSA is working.
Awesome! Are there any tutorials or scripts for analyzing the results of SparseDOSSA, other than https://github.com/himelmallick/BenchmarkMicrobiome?
That’s probably the best bet right now. We are however working on an updated version of SparseDOSSA, with new functionalities and more detailed tutorials, that should come out in the next couple of months. So stay tuned if it’d be relevant to you.