CGAL users discussion list

[Nico Kruithof <>]

Hi Patrice,

If I use the Skin_surface_pdb_reader example, the 1yyf.pdb molecule
takes 1200Mb of memory for the mixed complex and 700Mb mesh
(polyhedron). I didn't try subdividing the mesh. The program finished in
3m50s, but I have 2Gb of memory and it did not swap. The mesh is 115Mb
on disk.

About growing the atom sizes.
The skin surface is homeomorphic to the union of the input atoms. For a
shrink factor equal to one, it is the union of the input balls. For a
lower shrink factor, the radius of the balls decreases and smooth
patches appear between the balls. In order to avoid the decrease of the
radius, one can increase the radius of the input balls such that, after
shrinking (by the skin surface definition), the balls have their
original size again. Due to the increase of the radius, the union of the
blown up balls might change topology (and hence its skin surface). This
could be the explanation of the topological changes you saw. I hope that
this is clear.

The make_skin_surface_mesh_3 function, has three optional parameters:
void make_skin_surface_mesh_3(Polyhedron_3 &p,
                              WP_iterator begin, WP_iterator end,
                              double shrink_factor=.5,
                              int nSubdivisions=0,
                              bool grow_balls=true)
The last parameter enables the blowing up of the input balls. For the
SES, I would use this growing and use the default shrink factor of .5
(although you can choose any value between 0 and 1). Note that this is
not exactly the SES as the skin surface connects the balls by
hyperboloids and not with toroidal patches. But it is a 'look-a-like'.

If you have more questions, it might be better to use the cgal-discuss list.

Bests,
Nico

Patrice Rondao Alface wrote:
> Hello Nico,
> 
> Thanks for the packages,
> your code works very well!
> I have only a problem with very large molecules (surch as 1yyf.pdb) for
> which your method eats my 1Gb of RAM and 2Gb of swap memory. Anyway, for
> small and medium moleculas it is quite fast :-)
> I have a basic question about Skin Surfaces. What shrink factor should I
> use in order to extract the SES surface (or a very close approximation
> of it) from the Van der Waals balls? It looks to me that the skin
> surface with .5 is a little thinner than the SES should be.
> Am I wrong?
> It also causes topological holes that I would not have expected.
> I know that varying the shrink factor can modifiy the topology of the
> skin surface, but do you have a parameter that gives the SES surface? I
> wonder whether I have to multiply the van der waals radius of the
> spheres by another factor in order to match the skin surface and SES.
> 
> Could you please shed some light on this?
> 
> Best!
> 
> Patrice
>> Patrice,
>>
>> You can download the latest internal release from:
>> www.jive.nl/~kruithof/CGAL-3.3-I-287.tar.gz
>> and the documentation from:
>> www.jive.nl/~kruithof/Skin_surface_3.pdf
>>
>> I will remove the files after you have downloaded them. So, can you
>> please send me an e-mail after you did so? If you have any question or
>> suggestion with respect to the package, please let met know.
>>
>> Bests,
>> Nico
>>
>>
>> Patrice Rondao Alface wrote:
>>   
>>> Dear Nico,
>>>
>>> Thank you for your quick and detailed reply!
>>> I would be glad if you could send me the preliminary version.
>>>
>>> Best regards and thanks again,
>>>
>>> Patrice
>>>     
>>>> Dear Patrice,
>>>>
>>>> The code will be present in the new CGAL release that is due in a couple
>>>> of weeks. I can give you a preliminary version, if you want to start
>>>> right away. It can use a PDB reader that Daniel Russel wrote. For
>>>> citation, I would suggest:
>>>>
>>>> @article{DBLP:journals/comgeo/KruithofV07,
>>>>   author    = {Nico Kruithof and
>>>>                Gert Vegter},
>>>>   title     = {Meshing skin surfaces with certified topology.},
>>>>   journal   = {Comput. Geom.},
>>>>   volume    = {36},
>>>>   number    = {3},
>>>>   year      = {2007},
>>>>   pages     = {166-182},
>>>>   ee        = {http://dx.doi.org/10.1016/j.comgeo.2006.01.003},
>>>>   bibsource = {DBLP, http://dblp.uni-trier.de}
>>>> }
>>>>
>>>> Please let me know, if you have any questions about the package.
>>>>
>>>> Bests,
>>>> Nico
>>>>
>>>> Patrice Rondao Alface wrote:
>>>>   
>>>>       
>>>>> Dear Nico Kruithof,
>>>>>
>>>>> I am working on a research project of the Université catholique de
>>>>> Louvain in Belgium that aims at studying docking conditions between
>>>>> molecules. We are using VTK, CGAL libraries and we compute local
>>>>> features of the skin surface. We then analyze these features together
>>>>> with other physical measures coming from the PDB database to build a
>>>>> docking prediction model.
>>>>> Since a good representation (topologically correct and fast) of the skin
>>>>> surface is mandatory, we have found your excellent paper as well as an
>>>>> announcement of a new CGAL package for skin surface meshing on the CGAL
>>>>> webpage.
>>>>>
>>>>> Could you tell me whether this package (code or executable for linux) is
>>>>> available for academic research purposes? If so how could I get a
>>>>> version of it? Which paper would you like us to reference in our
>>>>> publication if we can make use of your code?
>>>>>
>>>>> Thanks in advance and best regards!
>>>>>
>>>>>     
>>>>>         
>>>     
> 
>