Bioinformatic tool -> berberine binding to enzymes

103 views
Skip to first unread message

Mega [Andreas Stuermer]

unread,
Aug 6, 2014, 9:27:00 AM8/6/14
to diy...@googlegroups.com
Hi! 

Could use some help with the following: 
One mode of action is that berberine and derivates activate cbl-ubiquitin ligase. Now I was told to identify structures where berberine binds to. So maybe to cbl, or some upstream protein.

I don't know of any bioinformatic tool that is advanced enough to solve this. Does anyone know of one? 

Best,
Andreas  

Aayush Aggarwal

unread,
Aug 6, 2014, 1:38:15 PM8/6/14
to diy...@googlegroups.com
Hello,

There is tool TRANSFAC which might help you out.
Also,there is list of tools at this link. http://www.geneinfinity.org/sp/sp_proteininteraction.html

I hope this would work.

Thanks
aayush aggarwal  


Patrik D'haeseleer

unread,
Aug 7, 2014, 3:09:30 AM8/7/14
to diy...@googlegroups.com
I would go to the Protein Data Bank, and look for protein structures that include berberine as a ligand: 

http://www.rcsb.org/pdb/results/results.do?qrid=C21BC50D&tabtoshow=Current 

Unfortunately, there's not that many of those, so you may also want to check some other berberine related compounds. Alternatively, you can start from the proteins that are associated with cbl-ubiquitin ligase, check what ligands those bind, and if there is something similar to berberine in there (there's a variety of metrics to measure how similar in structure two small molecules are). There may also be some other databases out there with known protein - small molecule interactions, that don't require a protein structure.

There's various pathway databases to retrieve genes that are in some way associated for your gene of interest. I like STRING, because it also includes gene coexpression data, genomic context, and co-citations:


Good luck!

Patrik

Mega [Andreas Stuermer]

unread,
Aug 7, 2014, 6:02:55 AM8/7/14
to diy...@googlegroups.com
Thanks a lot for your replies. It will take some time to try out all these tools ;) 

Mega [Andreas Stuermer]

unread,
Aug 7, 2014, 6:22:59 AM8/7/14
to diy...@googlegroups.com
Holy crap... Your link leads me to various links which probably all are search engines. And all those search engines have like 100 adjustments and give 1000s of results :O 

Mega [Andreas Stuermer]

unread,
Aug 7, 2014, 6:27:37 AM8/7/14
to diy...@googlegroups.com
There is probably no tool with one single search box where I write "berberine"  and another search box where I write "cbl-ligase" and it says "yes, it binds" or "no, it doesn't bind"? :D

Cathal (Phone)

unread,
Aug 7, 2014, 6:41:51 AM8/7/14
to diy...@googlegroups.com
I think you know the answer to.that question!


On 7 August 2014 11:27:37 GMT+01:00, "Mega [Andreas Stuermer]" <masters...@gmail.com> wrote:
There is probably no tool with one single search box where I write "berberine"  and another search box where I write "cbl-ligase" and it says "yes, it binds" or "no, it doesn't bind"? :D


--
Sent from my Android device with K-9 Mail. Please excuse my brevity.

digitalbio

unread,
Aug 7, 2014, 8:26:34 AM8/7/14
to diy...@googlegroups.com
You can do the same search at the NCBI, too.

Searching with "berberine" finds 61 structures, both proteins (mostly transcriptional regulators) and DNA.


PDB is nice, but I think the NCBI is a little easier to use.



On Wednesday, August 6, 2014 6:27:00 AM UTC-7, Mega [Andreas Stuermer] wrote:

Mega [Andreas Stuermer]

unread,
Aug 7, 2014, 9:26:53 AM8/7/14
to diy...@googlegroups.com
Hm ok. But bioinformatics is not advanced enough to make prediciton on ligand binding? I read that these methods are all empirical from previous studies

Mega [Andreas Stuermer]

unread,
Aug 7, 2014, 10:44:07 AM8/7/14
to diy...@googlegroups.com
I think I found one easy to use. Just gotta find the DB files.

http://hexserver.loria.fr/

Mega [Andreas Stuermer]

unread,
Aug 7, 2014, 11:01:24 AM8/7/14
to diy...@googlegroups.com
SDF file = PDB file?



On Wednesday, August 6, 2014 3:27:00 PM UTC+2, Mega [Andreas Stuermer] wrote:

Cathal Garvey

unread,
Aug 7, 2014, 11:05:06 AM8/7/14
to diy...@googlegroups.com
It's not a matter of bioinformatic advancement, it's a matter of
computational advancement. That sort of prediction assumes you can
physically compute things like protein folding, binding affinities and
such on a conventional computer, which is not presently possible to a
useful extent.

You can model proteins with computers, but they're not very good at it
and it takes ages to do. You could then try to predict binding
affinities of ligands to the proteins, but doing so would involve
trial-and-error computation and take a long time and be very error prone.

I'm not a computer scientist, but my intuition suggests that this sort
of problem might benefit from quantum computers. So, maybe in our
lifetime we'll have protein structure solvers and protein ligand
predictors that don't suck. But right now empiricism is the only
reliable way.
--
T: @onetruecathal, @IndieBBDNA
P: +353876363185
W: http://indiebiotech.com
0x988B9099.asc
signature.asc

David Murphy

unread,
Aug 7, 2014, 1:00:51 PM8/7/14
to diy...@googlegroups.com
I doubt it's in the class of problems that would benefit from quantum computing. 
The problem is that it's a subclass of the N-Body problem

 http://en.wikipedia.org/wiki/N-body_problem

Bigger/faster computers help a little but it costs O(N^2) where N is the number of particles you're simulating meaning that to simulate a slightly bigger protein or pair of proteins you'd need a vastly more powerful computer.

Add any slight errors in our models for subatomic particles. 
Add in simplifications depending on how low a level you want to simulate it.(Atoms? electrons/protons/neutrons?) 

I've seen high quality simulations of protein interactions where they simulate the forces between each atom but they take a long time and a huge amount of computational power. 

Unfortunately predicting how one 3d protein will bind to another protein is very hard. 

You may have been given a harder problem than your supervisor realized they were giving. 

I can't think of any easy way to approach your problem. 
Warning, semi-random musings, not recommended to try:
Possibly examine a selection of binding sites, pick out the common features important to the binding, convert each of them back to their sequences, compare to pick out the common parts that have high information content, come up with some kind of set of weighted masks then search for genes that contain sequences matching them. Though that's just off the top of my head so it's very likely to not work, have insurmountable problems, find an insane number of false positives, be a huge pain to even try and would still be computationally expensive but in a "I need a big computer" way rather than an "I need a computer the size of the planet" way.

Mega [Andreas Stuermer]

unread,
Aug 20, 2014, 5:32:36 AM8/20/14
to diy...@googlegroups.com
Found out how to do this:

Search Berberiene

http://www.ebi.ac.uk/pdbe-srv/pdbechem/chemicalCompound/show/BER

links -> downloaden pdb file (ideal), (representative)

(First I thought the ideal file would be best, but then recgonized that this does not neccessarily mean "ideal/perfect for analysis" but rather "ideal/simplified gas molecules". So the represdentative is probably better.

 

download receptor

http://www.rcsb.org/pdb/explore/explore.do?structureId=2Y1M

 

insert files .pdb. It compares the structures

http://hexserver.loria.fr/



Cathal Garvey

unread,
Aug 20, 2014, 5:37:05 AM8/20/14
to diy...@googlegroups.com
Fascinating, nice find! It seems it's built atop a tool called "hex"
that you can run locally. If you have a powerful computer handy, this
might help with your needs, because the hex server only accepts single
ligand:receptor pairs, rather than doing a BLAST-style search
(obviously, as doing this would be cripplingly expensive to do). By
running locally you could potentially set up a "ligand search engine",
although performing searches would likely take days (or more?) on
consumer hardware for a large enough protein set.

On 20/08/14 10:32, Mega [Andreas Stuermer] wrote:
> *Found out how to do this: *
> <http://hexserver.loria.fr/>
>

--
Twitter: @onetruecathal, @formabiolabs
Phone: +353876363185
Blog: http://indiebiotech.com
miniLock.io: JjmYYngs7akLZUjkvFkuYdsZ3PyPHSZRBKNm6qTYKZfAM
0x988B9099.asc
signature.asc

Mega [Andreas Stuermer]

unread,
Aug 20, 2014, 5:55:14 AM8/20/14
to diy...@googlegroups.com
Small Poblem...

I put the resulting file into Hex for visualization (did online analysis while downloading)

AAAnd this is what I get :/ An interactive 3D map. No binding value as in this paper

Anti-proliferative and metastatic protease inhibitory activities of protoberberines: An in silico and in vitro approaches
result cbl-berberine binding.png

John Griessen

unread,
Aug 20, 2014, 9:17:54 AM8/20/14
to diy...@googlegroups.com
On 08/20/2014 04:36 AM, Cathal Garvey wrote:
> By
> running locally you could potentially set up a "ligand search engine",
> although performing searches would likely take days (or more?) on
> consumer hardware for a large enough protein set.

You could also set hex up in a server image, rent a large memory server image
by the minute from amazon and pay for a search to run on much more hardware
faster. The rent might be just dollars per hour and run to completion in hours
instead of days.
Reply all
Reply to author
Forward
0 new messages