4
$\begingroup$

I'm currently doing an internship with my professor and we were using a few Protein-Protein Interaction Networks (PPIN) on a community detection algorithm MONET. The PPINs we used were unweighted, but we are curious how our results would change by adding weighted interactions between our proteins. I tried using the IntScore website (http://intscore.molgen.mpg.de) but unfortunately that's down. I also tried using some tools from the HIPPIE website, but the PPIN I am using has over 300,000 interactions which it can't process, even when using the API documentation.

Can anyone recommend any tools that I can use to get the edge weights for my interactions? I'll link an example of what I want my PPIN to look like. Thank you very much!

I am working with human data, specifically I am using NCBI gene entrez ID's as my main format for the algorithm I am using. We wanted the weights to represent the first one you mention, ie 'he reliability of the interaction prediction based on the experimental method used to predict it'. if experimental evidence isnt available for a given interaction, we would not mind an interaction score / confidence score given from text mining. If there is such a tool I can use, please let me know.

The first few lines are:

1785    26060   0.958126
1785    5747    0.999472
1994    2624    0.876802
988 1459    0.976516
7530    64837   0.0153325
5885    5901    0.997154
9318    122769  0.0171704
3091    7975    0.946822
3014    5351    0.464768
2099    10574   0.683967

Link to full network example: https://jumpshare.com/s/PScmwTMPTQTlDJGxf8fy

$\endgroup$
4
  • 2
    $\begingroup$ I don't know of any NCBI numerical ID system for proteins. Could those be NCBI gene IDs? I tried a couple and they seem to resolve to human genes, are you working with human data? $\endgroup$
    – terdon
    Commented Jul 9 at 9:02
  • 2
    $\begingroup$ Also, where would the weights come from? What do you want the weights to represent? Would it be the reliability of the interaction prediction based on the experimental method used to predict it? Would it be the likelihood of interaction? Or maybe the frequency of interaction? You can use weights for very many things, what do you want yours to be? $\endgroup$
    – terdon
    Commented Jul 9 at 9:04
  • $\begingroup$ Hey Terdon, thanks for responding. So i am working with human data, specifically I am using NCBI gene entrez ID's as my main format for the algorithm I am using. We wanted the weights to represent the first one you mention, ie 'he reliability of the interaction prediction based on the experimental method used to predict it'. if experimental evidence isnt available for a given interaction, we would not mind an interaction score / confidence score given from text mining. If there is such a tool I can use, please let me know. Thank you so much! $\endgroup$ Commented Jul 13 at 15:07
  • 1
    $\begingroup$ Please add this to your question so everyone can see it. I haven't worked on this kind of thing for more than a decade, but when I did, I made a list of the methods I found acceptable (I needed direct evidence of interaction) and used that. You need to know where each of the interactions you have came from and how they were derived. I expect you will need to build a new network. $\endgroup$
    – terdon
    Commented Jul 13 at 15:30

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.