sc.pl.heatmap show the wrong scale

[wangjiawen2013]

Dear, when I use sc.pl.heatmap(adata,markers,groupby='leiden',show_gene_labels=True) to show some marker genes, the colorbar value range is different from that generated by sc.pl.umap(). I am sure that umap show the expression value correctly and there must be something wrong with sc.pl.heatmap(), here are the images:

For example, the expression value of gene 'Sct' is 0-7.5, while it changes to 0-1e7 on the heatmap, the value is so high that it may scale other gene's expression value to almost zero and nothing can be seen on heatmap for other genes. Some other genes have the same issue. Do you know the reason ?

[LuckyMD]

Hi,
Just a guess... have you set use_raw to different values in both cases. It would be interesting to see what the max expression value is for 'Sct' in adata.X and adata.raw.X.

[wangjiawen2013]

Default parameter was set for use_raw in each cases. They were the same. 1e7 is a so large value that it is not proportional to 0, 1, 2, 3 4. I doubt that sc.pl.heatmap treats some genes with a certain name and expression value as special characters or exception values. In the above images, if I remove “Sct" from the marker gene list, then another gene which works well before the remove would has the similar issue ! Now look carefully at the above heatmap, you can see that at the bottom of the heatmap, Sct show a very high expression value (it's very thin perhaps you cannot see it).

[LuckyMD]

I did not see the 1e7. It seems that the heatmap colormap scale is just off then. Not sure how to change that though... @fidelram?

I'm not sure what I'm meant to be looking for that is very thin otherwise...

[fidelram]

This is strange and I had not seen it before. Maybe something related to a newer version of matplotlib?

You can fix the problem by doing sc.pl.heatmap(...., vmax=7).