[discuss ifpsc] Re: Feedback requested re: 5th IFPSC proposal
I think Andreas makes the critical points. Knowledge-based methods are well-established in this field. Within the knowledge domain the accuracy is typically 0.3, more generally it is more like 0.9. Some pharma companies do not even both to measure LogP for octanol, relying on cLogP. Furthermore, two compounds will tell you nothing statistically. For comparison, for those interested, our blind challenge (SAMPL: http://sampl.eyesopen.com) has a water-vacuum component of 63 compounds of varying complexity, some even drug-like. 

Anthony


On Feb 5, 2008, at 7:57 AM, Elliott Jr,J. Richard wrote:


Andreas makes a very good point.  We sacrifice something in terms of
ready recognition of the property we are predicting, however. Is there 
some other property, perhaps related to ADME properties of
pharmaceuticals that might be a bit less fully developed?  This would
give the best of both worlds.

I still have a doubt about Jonathon's concern regarding the number of
compounds.  In the measurements that I have made, the hard part is
getting the apparatus all fresh and re-calibrated.  Once that is done,
it is almost a waste to measure just two things. Maybe the price picks up if you go from five to ten, but I don't see how the price can be very different for two or five. Having more compounds reduces the element of 
winning by wild luck, also mentioned by Andreas.  Maybe you like the
element of wild luck, but I think it undermines the seriousness of the
effort.

JRE

-----Original Message-----
From: discuss-ifpsc-bounce@xxxxxxxxxxxxx
[mailto:discuss-ifpsc-bounce@xxxxxxxxxxxxx] On Behalf Of Andreas Klamt
Sent: Tuesday, February 05, 2008 8:13 AM
To: info@xxxxxxxxxxxxxxxxxxx
Cc: discuss-ifpsc@xxxxxxxxxxxxx; fluidproperties@xxxxxxxxxxxxx
Subject: [discuss ifpsc] Re: Feedback requested re: 5th IFPSC proposal

Dear colleagues,

thanks for working out the proposal. Basically we would be happy to
participate in such contest. Nevertheless, I am afraid that
octanol-water partition coefficients are not an optimal target property, because there is so much experimental data published, and because there are such good increment methods available as CLOGP. You will have a hard time finding example molecules which are strongly mispredicted by CLOGP 
an which are nevertheless simple enough to be treated by force-fields.
Hence participants would have the predicted CLOGP values as de-facto
experimental values available and then could tune their methods until
they achieve results close to those of CLOGP. Thus they would be quite
sure that they are no more off than about 0.3 log units, which anyway
will be the limit any of the methods can realstically expect to achieve 
(obviously, accidentally one can be better).

In this way I am afraid that the contest looses its blind test
character. Why dont you measure the partition coeff. in a different
system, e.g. water-octylaldehyde, or water- benzylalcohol, or similar.
From the modelling perspective this would be of comparable complexity,
but since no good simple increment systems are available for such, one
could be sure that the submissions are blind modelling results.

Sincerely

Andreas

IFPSC schrieb:

Dear colleagues,

We're making plans for the 5th Industrial Fluid Properties Simulation
Challenge in 2008.  We have posted a problem proposal here:

http://fluidproperties.org/proposal-5th-simulation-challenge

We request that you take a look at the proposal and send feedback and
suggestions for improvement to ifpsc@xxxxxxxxxxxxxxxxxxx
<mailto:ifpsc@xxxxxxxxxxxxxxxxxxx> by Feb 15.

(Apologies if you receive this email more than once due to overlap
between the different email list).

Sincerely,
The IFPSC Steering Team





--
------------------------------------------------------------------------
-----
Dr. habil. Andreas Klamt
COSMOlogic GmbH&CoKG
Burscheider Str. 515
51381 Leverkusen, Germany

Tel.: +49-2171-73168-1  Fax:  +49-2171-73168-9
e-mail: klamt@xxxxxxxxxxxxx
web:    www.cosmologic.de
------------------------------------------------------------------------
-----
COSMOlogic
      Your Competent Partner for
      Computational Chemistry and Fluid Thermodynamics
------------------------------------------------------------------------
-----



========================================================

The Industrial Fluid Properties Simulation Collective is an
international collaboration between industry, academia, and national
labs aimed facilitating and promoting the use of molecular modeling and 
simulation methods for the prediction of fluid properties.

discuss-ifpsc@xxxxxxxxxxxxx is an email list for discussing issues
related to the Industrial Fluid Properties Simulation
Collective...fluidproperties.org

If you are not subscribed, you can do so by sending email to
discuss-ifpsc-request@xxxxxxxxxxxxx with 'subscribe' in the Subject
field

You can unsubscribe by sending email to
discuss-ifpsc-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject.

These and other types of changes (switch to digest mode, get RSS feed,
etc.) can also be made here:
http://www.freelists.org/list/discuss-ifpsc

Online, searchable archives of this list are available at
http://www.freelists.org/archives/discuss-ifpsc

========================================================
The Industrial Fluid Properties Simulation Collective is an international collaboration between industry, academia, and national labs aimed facilitating and promoting the use of molecular modeling and simulation methods for the prediction of fluid properties.
discuss-ifpsc@xxxxxxxxxxxxx is an email list for discussing issues related to the Industrial Fluid Properties Simulation Collective...fluidproperties.org
If you are not subscribed, you can do so by sending email to discuss-ifpsc-request@xxxxxxxxxxxxx with 'subscribe' in the Subject field
You can unsubscribe by sending email to discuss-ifpsc-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject.
These and other types of changes (switch to digest mode, get RSS feed, etc.) can also be made here: 
http://www.freelists.org/list/discuss-ifpsc

Online, searchable archives of this list are available at 
http://www.freelists.org/archives/discuss-ifpsc



========================================================

The Industrial Fluid Properties Simulation Collective is an international 
collaboration between industry, academia, and national labs aimed facilitating 
and promoting the use of molecular modeling and simulation methods for the 
prediction of fluid properties.

discuss-ifpsc@xxxxxxxxxxxxx is an email list for discussing issues related to 
the Industrial Fluid Properties Simulation Collective...fluidproperties.org

If you are not subscribed, you can do so by sending email to  
discuss-ifpsc-request@xxxxxxxxxxxxx with 'subscribe' in the Subject field

You can unsubscribe by sending email to discuss-ifpsc-request@xxxxxxxxxxxxx 
with 'unsubscribe' in the Subject.

These and other types of changes (switch to digest mode, get RSS feed, etc.) 
can also be made here:
http://www.freelists.org/list/discuss-ifpsc

Online, searchable archives of this list are available at 
http://www.freelists.org/archives/discuss-ifpsc

Other related posts: