Pypath and direct and indirect interactions

[Marcio Luis Acencio]

Dear all,

I am interested in using pypath to retrieve only direct causal interactions between entities from literature curated resources such as Signor and SignaLink3.

However, some resources, like the ones cited above, have both direct and indirect causal interactions. Is it possible to get, for example, only direct causal interactions from these resources using pypath?

One more question: I have noticed that via Omnipath it is possible to obtain about 5000 interactions from Signor; however, this database - at least the 2015 version - has 11000 interactions. Why this difference?

One last question: I have accidentally found one protein-protein interaction (O60477 _Q92484) in which, according to Omnipath, O60477 stimulates/activates Q92484. I could retrieve this information only directly from the paper since nor STRING neither HPRD seem to show this information. Moreover, this stimulation/activation is via gene expression. So, please, let me know how this causal interaction has been assigned. This is important because maybe there are similar cases in Omnipath.

Best,

Marcio

_________________________

Marcio Luis Acencio, PhD

Postdoctoral Fellow

Department of Cancer Research and Molecular Medicine

Norwegian University of Science and Technology (NTNU)

Telephone: +47 72 82 6925

e-mail: marcio.l...@ntnu.no

[Dénes Türei]

Dear Marcio,

Thanks for your mail.

This is indeed important and I will check your example in detail and
give explanation a bit later.

In OmniPath we defined the default input parameters for all data
sources with the intention to include only direct interactions.
However depending on the source and format it might be not possible or
not perfect, this is the way how a number of indirect interactions
might sneak into OmniPath.

Signor besides direct protein-protein interactions contains a number
of other kinds of interactions, e.g. small molecules, transcription,
indirect, etc. These are all discarded when we load PPIs. A small
number might get lost due to ID translation issues (not likely at
Signor), and also bidirectional or redundant interactions going to be
assigned to one edge in the undirected igraph representation of
pypath.
The undirected igraph object can be converted to a directed one using
the `get_directed()` method of PyPath.
As igraph itself has no signed (+/- ) representation, these data are
handled by the Direction objects assigned to each edge, e.g:

pa.graph.es[0]['dirs']

These objects have many methods to retrieve direction and sign
information, unfortunately not all documented, but I will do my best
to update it: http://pypath.omnipathdb.org/#pypath.main.Direction
You can iterate over the edges and using the methods above get the
outcome most suitable for you (for example you can get the majority
vote from all resources, or get all signs from any resource, or delete
the edges having no sign, etc).

If your platform for further analysis is not Python, you can simply
export the data by export_tab or export_sif methods.

I hope this helps, tell me if you have further questions.

Best,

Denes

> --
> You received this message because you are subscribed to the Google Groups
> "Omnipath" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to omnipath+u...@googlegroups.com.
> To view this discussion on the web, visit
> https://groups.google.com/d/msgid/omnipath/1493763246092.76540%40ntnu.no.
> For more options, visit https://groups.google.com/d/optout.



--
Denes Turei, Ph.D.

postdoc @ EMBL
+4915166299976 Germany
+447442970610 UK
de...@ebi.ac.uk
http://www.ebi.ac.uk/~denes
public key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x5706A4B609DD65A6

[Dénes Türei]

Dear Marcio,

In this file: http://string-db.org/download/protein.actions.v10.5/9606.protein.actions.v10.5.txt.gz

We find the row:

9606.ENSP00000265922    9606.ENSP00000357425    activation    activation    1    800

Here the existence of the interaction came from HPRD and it was kept in OmniPath because STRING described it as a stimulatory, directed interaction.

Since then I removed STRING from the list of those resources which I get direction/sign information from when building OmniPath. I found that the file cited above contains many mutual bidirectional links, and possibly these are just interactions with unknown direction.

Because of this the interaction you cited is not in OmniPath any more, and will disappear from the webservice once I update the data also there.

Thanks again for the feedback!

Best,

Denes