Inconsistency between calculated RSS scores and dotplotheatmap

[laurajohanne]

Hi,

I am having som inconsistency when comparing which eRegulon specificity when RSS scores and dotplotheatmap based on "direct_gene_based_AUC" and "direct_region_based_AUC".

It appers that some eRegulons should be specific to a certain cell type but when comparing to the dotplotheatmap specifically the color ("direct_gene_based_AUC") it does not show similar results.

e.g. The RSS scores for the eRegulon Atf3_direct_+/+_(239g):

rss = regulon_specificity_scores( scplus_mudata = scplus_mdata, variable = "scATAC_counts:celltype", modalities = ["direct_gene_based_AUC"] )

rss.filter(like='Atf3')

and based on the average AUC value for the eRegulon Atf3_direct_+/+_(239g):

Code for making the dotplotheatmap;
heatmap_dotplot( scplus_mudata = subset_mdata, color_modality = "direct_gene_based_AUC", size_modality = "direct_region_based_AUC", group_variable = "scRNA_counts:celltype", eRegulon_metadata_key = "direct_e_regulon_metadata", color_feature_key = "Gene_signature_name", size_feature_key = "Region_signature_name", feature_name_key = "eRegulon_name", sort_data_by = "direct_gene_based_AUC", orientation = "vertical", figsize = (16, 20) )

The color_matrix:

group_by = scplus_mdata.obs["scRNA_counts:celltype"].tolist()

color_features = scplus_mdata.uns["direct_e_regulon_metadata"]["Gene_signature_name"].unique()

color_matrix = direct_gene_based_AUC[color_features]

color_matrix_avg = color_matrix.groupby(group_by).mean()

color_matrix_avg.filter(like='Atf3')

Since both are calculated on the same dataframe; scplus_mdata.mod["direct_gene_based_AUC"], am I not quite sure why I get this inconsistency...

Hope you can help.

Best regards,

Laura

[SeppeDeWinter]

Hi @laurajohanne

Both are different metrics:

The RSS in essence calculates how similar the actual distribution of AUC values for your eRegulon is to what the distribution would be if the regulon is only active in a single cell type.

While the values in the dotplotheatmap are average AUC values per cell type.

I hope this helps?

Best,

Seppe

[laurajohanne]

Thank you, that helps.

Best regards, Laura.

[YH-Zheng]

Hi @SeppeDeWinter,

I would like to know how to filter cell type specific eRegulons by RSS value? Is it to select a threshold value for filtering?

Best regards,
Yuhui