[Dock-fans] Documentation of Dock 6.2 test suite results

[Scott Brozell]

Hi,

On Sat, 17 May 2008, James Parker wrote:

>   I have recently attempted to install DOCK 6.2 and run the test suite 
> on a Fedora Core 6 system.  The install compiled simply once I installed 
> g77, and I was able to run the testsuite as well.  At the conclusion, I ran
>  > make check
> as described in the user manual and there were several differences 
> reported. 

>   I was hoping someone could point me to a website/documentation that 
> would enable me to determine if these differences are significant or 
> not.  I would think they are since the reported differences are 
> generally related to "score" values that seem to be integers that are 
> significantly different, as opposed to "real" numbers that differ in the 
> rounding of trailing digits.  The differences I found were not listed in 
> the KNOWN_FAILURES file.
> 
> I did a quick check through the last few months of mail-list archives 
> and did not see anything.  Any help would be appreciated.

See
http://blur.compbio.ucsf.edu/pipermail/dock-fans/2008-March/001488.html

Note that the output files are filtered to reduce floating point noise:
./primary_secondary_score/secondary5.dockout.dif
108c108
<  Conformations:    17
---
>  Conformations:    19

These are integers.

110c110
<  Contact Score:          148.
---
>  Contact Score:          173.

These are truncated floating point numbers.


http://blur.compbio.ucsf.edu/pipermail/dock-fans/2008-May/001583.html
Recent work suggests that computer hardware,
especially 32bit vs 64bit, may be the dominant part of the standard blurb:
The sources of such differences are frequently platform dependencies
from computer hardware, operating systems, and compilers that impact
arithmetic precision and random number generators.

Scott