Direction reversal

[hitwhicks]

Hello!

I am using scVelo with a dataset where I know the terminal clusters ('9','8','6','7'). The tool works great, except it predicts a direction reversal like in a lot of previous cases! I tried running all three models and I get the same result with the velocity directions going from terminal to presumably root clusters.

Here are my top 10 velocity genes:

print(adata.var['velocity_genes'].sum(), adata.n_vars)
69 771

top_genes = adata.var['fit_likelihood'].sort_values(ascending=False).index[:300]
scv.pl.scatter(adata, basis=top_genes[:10], frameon=False, ncols=5,color='seurat_clusters',save='velocity_genes.png')

I also looked at the unspliced fractions like in certain other examples like the Erythroid progenitor case:

counts_s = scv.utils.sum_var(adata.layers['spliced'])
counts_u = scv.utils.sum_var(adata.layers['unspliced'])
fractions_u = counts_u / (counts_s + counts_u)
scv.pl.scatter(adata, color=fractions_u, smooth=True,save='fractions.png')

other than knowing the terminal clusters I will admit not much is known about biology behind the cell type transitions in my case. So I wanted to get your opinion before I went ahead and interpreted it!

Thanks!

[hitwhicks]

Hello!

I am using scVelo with a dataset where I know the terminal clusters ('9','8','6','7'). The tool works great, except it predicts a direction reversal like in a lot of previous cases! I tried running all three models and I get the same result with the velocity directions going from terminal to presumably root clusters.

Here are my top 10 velocity genes:

print(adata.var['velocity_genes'].sum(), adata.n_vars)
69 771

top_genes = adata.var['fit_likelihood'].sort_values(ascending=False).index[:300]
scv.pl.scatter(adata, basis=top_genes[:10], frameon=False, ncols=5,color='seurat_clusters',save='velocity_genes.png')

I also looked at the unspliced fractions like in certain other examples like the Erythroid progenitor case:

counts_s = scv.utils.sum_var(adata.layers['spliced'])
counts_u = scv.utils.sum_var(adata.layers['unspliced'])
fractions_u = counts_u / (counts_s + counts_u)
scv.pl.scatter(adata, color=fractions_u, smooth=True,save='fractions.png')

other than knowing the terminal clusters I will admit not much is known about biology behind the cell type transitions in my case. So I wanted to get your opinion before I went ahead and interpreted it!

Thanks!

@WeilerP Any input on this? Sorry just very curious! :)

[WeilerP]

@hitwhicks, please have a look at e.g. #216 or #462. Your phase portraits simply do not have the expected shape s.t. dynamics cannot be inferred (confidently).