how does sc.queries.enrich handle up- and down-regulated genes

How does sc.queries.enrich handle upregulated and downregulated differentially expressed genes? Are they both input into GProfiler, with no distinction made between which are up and which are down?

I ask because it's important for interpretation. For example, if both upregulated and downregulated genes are input to GProfiler without distinction, then if rank_genes_groups had found all downregulated genes for phenotype A, then the pathways reported for phenotype A would actually be enriched in phenotype B.

My current understanding is that all genes are passed together. If you supply a min log2fc_min > 0, it will include only upregulated genes, but otherwise it will include all. Is this correct?

More generally, is there some place I could view the API code, to get a better sense of how this function works? On GitHub all I can see is gprofiler = GProfiler(user_agent="scanpy", return_dataframe=True), and I can't find the details in GProfiler's documentation either. Where is the "container" object created?

p.s. I think it might be helpful to clarify the syntax for passing parameters to gprofiler_kwargs. It took some playing around for me to find the right combination of string + boolean for gprofiler_kwargs={'no_evidences':False}

Thanks!

Start by locating the implementation and API entry point for sc.queries.enrich, then trace how rank_genes_groups results and gprofiler_kwargs reach GProfiler. Document whether up- and downregulated genes are distinguished, how min_log2fc_min affects inputs, where the container is created, and how the no_evidences option is passed.