different mode produces the same outputs in dyn.vf.rank_jacobian_genes

[chansigit]

Dear authors,

I tried 'full reg'/ 'full eff' modes in the dyn.vf.rank_jacobian_genes() function

full_reg_rank = dyn.vf.rank_jacobian_genes(abm2,
                                           groups='ann0608',
                                           mode="full reg",
                                           abs=False,
                                           output_values=True,
                                           return_df=True)
full_eff_rank = dyn.vf.rank_jacobian_genes(abm2,
                                           groups='ann0608',
                                           mode='full eff',
                                           abs=False,
                                           exclude_diagonal=True,
                                           output_values=True,
                                           return_df=True)

But I got the same dataframes

are they supposed to be the same under the two modes? or did i misoperate anything when computing the jaccobian?

p.s.

I calculated the jacobian like this:

dyn.pp.top_pca_genes(abm2, n_top_genes=100);
top_pca_genes = abm2.var.index[abm2.var.top_pca_genes];

# 加一些你认为重要的
top_pca_genes =[
    'Nfatc1','Nfatc2','Nfatc3','Nfkb1','Rbpj','Rela','Stat5a',
    'Stat5b','Stat6','Jun','Maml1','Maml2','Maml3','Nfatc1','Nfatc2','Nfatc3',
    'Rbpj','Rela','Stat1','Stat4','Mapk11','Mapk13','Mapk14','Mtor','Npas2',
    'Runx1','Smad2','Stat3','Smad4','Smad3','Rorc','Gata3']+[
    "Abcb11","Abcc2","Abcc3","Abcc4","Abcd3","Aqp8","Aqp9","Atp8b1","Ceacam1",
    "Ceacam2","Cyp7a1","Fgf15","Gm5724","Gm6614","Hnf1a","Mip","Nr0b2","Nr1h4",
    "Prkaa1","Slc10a1","Slc10a2","Slc10a3","Slc10a4","Slc10a4-ps","Slc10a5","Slc10a6",
    "Slc10a7","Slc51a","Slc51b","Slco1a1","Slco1a4","Slco1a5","Slco1a6",
    "Slco1b2","Slco1c1","Slco2b1"]+ list(top_pca_genes)

top_pca_genes=list(set(top_pca_genes))

dyn.vf.jacobian(abm2, regulators=top_pca_genes, effectors=top_pca_genes);

thank you.

[Xiaojieqiu]

Thanks for reporting this important bug. My latest commit should fix this bug. Please see this line of code, the J should be transposed: https://github.com/aristoteleo/dynamo-release/blob/f26039c7068bb00819b9a88805aeac73d2ff3af9/dynamo/vectorfield/vector_calculus.py#L1784

Please let me know whether the result after the fix makes sense

[chansigit]

Thank you Xiaojie. It has been fixed.