Global degree does not change when changing significant edge value

[MeeshCompBio]

The global degree of genes are always representative of a z-score of 3, it should update when you change the significant edge score.

In [69]: cob = co.COB("W2691")                                                                                                                                                                                     [LOG] Mon Feb  5 16:33:41 2018 - Loading Expr table                                                                                                                                                                [LOG] Mon Feb  5 16:33:41 2018 - Building Expr Index                                                                                                                                                               [LOG] Mon Feb  5 16:33:41 2018 - Loading RefGen                                                                                                                                                                    [LOG] Mon Feb  5 16:33:41 2018 - Building Indices                                                                                                                                                                  [LOG] Mon Feb  5 16:33:41 2018 - Loading Coex table                                                                                                                                                                [LOG] Mon Feb  5 16:33:44 2018 - Loading Global Degree                                                                                                                                                             [LOG] Mon Feb  5 16:33:44 2018 - Loading Clusters                                                                                                                                                                  [LOG] Mon Feb  5 16:33:44 2018 - Building Indices
In [70]: Gene = cob.refgen.from_id("TRAESCS3D01G246000")
In [71]: cob.global_degree(Gene)
Out[71]: 120
In [72]: len(cob.neighbors(Gene, return_gene_set=True))
Out[72]: 120 
In [73]: cob.set_sig_edge_zscore(2.5)
In [76]: cob.global_degree(Gene)
Out[76]: 120
In [77]: len(cob.neighbors(Gene, return_gene_set=True))
Out[77]: 902

[schae234]

This is a mess right now, the state of the z-score threshold is not well described. Question: Should updating the z-score threshold influence the underlying database at all? I.e. if you set z-scrore to 4 should that modify the coex database table so there is one point of truth? If another instance of the network is started, should it see that the z-score cutoff is 4? This seems easiest to implement but it also restricts concurrency, I've run analyses where I've had 4 different instances of camoco running to test the effects of cutoffs. Ideally we would be able to do this, but it makes the implementation a little harder!

I think we should have two cutoffs. One that gets calculated during co-expression and another that you can slide around and only affect the current instance of the network. Thoughts?

[MeeshCompBio]

I agree with your idea, I don't think updating the Z-score should affect the underlying database. There should at least be documentation that global degree is based off the DB cutoff an not the current threshold set. That being said, maybe we would add another arg into the global degree function to calculate based off DB or current instance?

On Sat, Aug 18, 2018 at 7:44 AM Rob Schaefer notifications@github.com wrote:

This is a mess right now, the state of the z-score threshold is not well described. Question: Should updating the z-score threshold influence the underlying database at all? I.e. if you set z-scrore to 4 should that modify the coex database table so there is one point of truth? If another instance of the network is started, should it see that the z-score cutoff is 4? This seems easiest to implement but it also restricts concurrency, I've run analyses where I've had 4 different instances of camoco running to test the effects of cutoffs. Ideally we would be able to do this, but it makes the implementation a little harder!

I think we should have two cutoffs. One that gets calculated during co-expression and another that you can slide around and only affect the current instance of the network. Thoughts?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/LinkageIO/Camoco/issues/81#issuecomment-414055533, or mute the thread https://github.com/notifications/unsubscribe-auth/ANr0AJ_PUMt_Z6JX9AIIVsk1PT9_Cjbvks5uSAw8gaJpZM4R6Nn2 .

-- Jean-Michel Michno (Meesh) Bioinformatician University of Minnesota mich0391@umn.edu 952-212-6188