Re: [Scaffoldhunter-devel] GSOC 2014 - Scaffold Tree Layout

[Till S. <til...@tu...>]

Hi, 
first of all: please ask your questions over the mailing list, so we do not need to repeat our self and the other students also profit from the discussion. 

Am Montag, 17. März 2014, 19:26:19 schrieben Sie:
> Hi Till,
> Thanks for your reply , 
> i have questions about Scaffold Tree Layout idea , during last two days i read many papers about visualization of large graphs , i think that we could add  fisheye layout [1] or a force-directed layout algorithm named OpenOrd [2] . what's your opinion ? Is this what you think about ?
Thanks for your investigations. I think the two layouts mentioned, are not suitable for the tree layout, as they are for general graphs and it is much easier to layout trees. Furthermore, the depth of a node in the scaffold tree is very important for the user, as it indicates the size (number of rings) of the scaffold. Therefore the levels should be clearly visible in an optimal layout. The problem with the current radial tree layout is mainly  a good space utilization for large datasets. If you have 50k molecules, organize them in the scaffold tree and zoom to the overview, you see just as small ring of the tree. The radial layout has not much advantage over a linear layout in this situation. You can download a large sdf file from [1].

[1] https://ls11-www.cs.tu-dortmund.de/people/schaefer/publication_data/ks_compound.sdf.zip

Greetings 
Till 

> 
> 
> 
> [1] http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=1382906[2] http://gephi.org/2010/openord-new-layout-plugin-the-fastest-algorithm-so-far/
> > From: til...@tu...
> > To: sca...@li...
> > CC: aly...@ho...
> > Subject: Re: [Scaffoldhunter-devel] GSOC 2014 -  Sub graph network
> > Date: Mon, 17 Mar 2014 12:52:55 +0100
> > 
> > Hi, 
> >  
> > Am Samstag, 15. März 2014, 21:57:53 schrieb Aly Osama:
> > > Hello, I am a pr-junior student at Computer Engineering and Software Systems  .i have good experience with java and GUI with Swing library for 2 years , i have worked on a research project at AinShames university and develop a graphical tool that will used in VLSI industry using swing for 5 months . i am also interested in problem solving  and participate in Local ACM competition ACM ECPC . i have also some experience with image processing using MATLAB and Sci-kit library with python in Scanning Electron Microscopy Image Analysis software .
> > thank you for your introduction.
> > 
> > > I am really interested in Scaffold Hunter project as want to contribute in this field  so i will participate in this project for GSOC 2014 , i have got an overview of the user side in addition to getting in touch with the code , and i start learning Piccolo2D for 3 days . i think that  i can participate at Sub-graph Network or Molecule Clouds , 
> > As we have some Molecule Cloud proposals already, it will improve your chances to get accepted if you go for the subgraph network.
> > 
> > 
> > > I will work further on these ideas to choose one of them , but could you give me more details about sub-graph network ? 
> > The scaffold principle is a classification of molecules based on their (ring) graph structure. If you cut all the side chains (that means paths and trees, which are not part of a circle) of a molecule you get a so called murcko scaffold. As different Molecules might have the same murcko scaffold, you can think of a class of molecules with the same murcko scaffold. This scaffold itself serves as a descriptor of this class.
> > 
> > To take this approach a step further, one can select a rings of the scaffold and remove this ring. Therefore, you get a smaller parent scaffold. If you select the this ring with some strategy, you get a hierarchy of scaffolds and a hierarchical classification scheme for your molecules. This is what the scaffold tree view visualizes. 
> > 
> > Now, one can imagine, the the selection of the ring to remove can be done in various ways. That means with different ring selection strategies, you will get different classification hierarchies. 
> > 
> > The scaffold / subgraph network, is a DAG, where each scaffold has as many parents as there exists rings. That means for each ring, there exist one parent (the one which is created by removing this ring). All possible scaffold hierarchies can be embedded into this graph. 
> > The goal here is to find a layout for this DAG and also some filter techniques, because the DAG can be dense / huge. 
> > 
> > The provided papers are describing this in more detail. 
> > 
> > Greetings,
> > Till
> >  	   		  
>  		 	   		  
-- 
Dipl.-Inf. Till Schäfer
TU Dortmund University
Chair 11 - Algorithm Engineering
Otto-Hahn-Str. 14 / Room 237
44227 Dortmund, Germany

e-mail: til...@cs...
phone: +49(231)755-7706
fax: +49(231)755-7740
web: http://ls11-www.cs.uni-dortmund.de/staff/schaefer
pgp: https://keyserver2.pgp.com/vkd/SubmitSearch.event?&&SearchCriteria=0xD84DED79