# How many proteins could participate in a complex

Disclaimer: I’m a computer science student with minimum knowledge of biology.

I’m working on an algorithm to cluster proteins in Protein-Protein-Interaction Networks to find protein-complexes. While working on that I stumbled upon the question how many different proteins can be part of a protein complex. (I'll call this the size of the complex from this point.)

I started by counting the the participants from all Corum complexes. I got sizes ranging from 1 to 143:

[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 22, 23, 24, 25, 26, 27, 28, 30, 31, 32, 33, 36, 37, 38, 40, 44, 45, 47, 48, 62, 68, 78, 80, 104, 143]

The distribution is skewed to smaller sizes with 3 participants counted 1465 times and most of the bigger sizes from around 30 counted 1 or 2 times.

{44: 1, 36: 1, 32: 1, 47: 1, 78: 1, 48: 1, 31: 1, 143: 1, 40: 1, 26: 1, 38: 1, 62: 1, 104: 1, 23: 1, 20: 2, 22: 2, 33: 2, 80: 2, 37: 2, 45: 2, 28: 2, 68: 2, 27: 2, 30: 3, 19: 3, 24: 4, 25: 4, 18: 6, 17: 11, 15: 19, 1 6: 21, 14: 23, 11: 25, 12: 28, 13: 30, 10: 55, 9: 57, 8: 72, 7: 83, 6: 100, 1: 127, 5: 237, 4: 499, 2: 1370, 3: 1465}[Sorry for not sorting...]

My main question from this first dip into data is, are there any assumptions to make about the size of complexes? Are these big complexes for example special cases and normally complexes are limited to a size of around n? Is there maybe even an upper limit of participants in a complex?

Anything would be helpful for me to minimize runtime.

• This is suggestion as to how you could clarify the question to help us help you, without your having to resort to socializing, as per my answer. Could you sort the non-redundant list by number of participants and then present us with say 20 entries that span your putative cut-off, just showing the number of participants and the name of the complex? That way the question becomes, can we apply some different descriptor to complexes above the cut-off, which you can use in your report. Btw. I downloaded one of the files. I would advise having only one instance of a complex, e.g. not mice and man. Commented May 28, 2020 at 11:34
• Hi. At first thanks for the answer. I can do that, but it'll probably take me some time as I'm right now working at several different spots concerning this project.
– obvg
Commented May 28, 2020 at 11:59
• [Was to slow to edit...] Still I want to clarify more what my project is. The Idea is can we find protein complex candidates just by looking at the structure of PPI's. I'm especially working with quasi-cliques. So I try to find a configuration where my algorithm finds all known complexes and more. These extra findings are then seen as candidates. That's my reasoning for asking this question. The project itself is more like "Let's see what happens and if we want to further work in this direction" Also it's my bachelor thesis, so the scope is definitively limited.
– obvg
Commented May 28, 2020 at 12:10