Questions of 3D QSAR Pharmacophore Generation Result

Q1.
In the result, the Null cost value is too close to the total cost value.
How to increase Null cost value ?

Q2.
If the config value > 17, how to decrease the config value ?

Q3.
In some reference papers, they have calculated correlation value for the testing set using Catalyst.
But, I don't know how to calculate correlation in DiscoveryStudio.
So the first one question is how to calculate the correlation value for the testing set in DiscovreyStudio and
another one question is could you please give me the correlation formula you used in DiscoveryStudio?

Thank you for your kindly help!
Answers 4
SB
SB 2010-06-10

Hai K.C,

Q1:How to increase Null cost value ?
A: Make sure that your training set with structure and activity (4-fold)diversity. Some times, the combination of pharmacophore features also responsible cost difference.

Q2:If the config value > 17, how to decrease the config value ?

A: Please check the below link for response from Jon Sutter (Accelrys) regarding Configuration cost:

https://community.accelrys.com/message/3766

Q3:For Test set correlation?
A: Prepare the test set molecules by adding attributes (Activ, Uncert) and generate conformations using diverse conformation search and submit for Ligand pharmacophore mapping option in protocols/Pharmacophore module. Then open the output.sd file and check datatable, it will create two fields with Est and Fit value. Generate chart/point plot between Activ vs Est, Then in display options, select R2 Value.

Suneel Kumar BVS

HP
HP 2009-09-24
Hi, Suneel Kumar BVS,

Very thank you for your suggestion.

In Q1,
I have followed your suggestion, but the Null cost value stall too close to the total cost value.
Have any strategy which can increase the values(null cost - total cost)?

In Q3,
I can't find the display options to generate R2 value at last step of your suggestion.
Could you explain the detial of generating R2 value?

The following is what I do:
step1. I open the output.sd file.
step2. I select two fields, Est and Fit value.
step3. Generate point plot from chart on a new Tab.
step4. There is no "display options", but just only "Display Style".

Thank you for your kindly help!
SB
SB 2009-09-24
Hai K.C

For test set correlation:
Try this options:
step1. open the output.sd file.
step2. Generate plot from chart/Point plot between Activ and Est.
step3. Right click on plot then, Add/Remove/trendline/selectall
and enter You will get R2 Value.

Suneel Kumar BVS
comment imagecomment image
HP
HP 2009-09-24
Hi, Suneel Kumar BVS,

Very thank you for your suggestion.
I have question about R2 value.
When I use "3D QSAR Pharmacophore Generation", the correl=0.958354(Hypothesis 1).
When I followed your instructions to recompute R2 value using Point plot, I get R2 value = 0.497.
But, two fields Est and Fit value are from Hypothesis01 output sd file(Hypothesis 1)
Why are the two values different?
Dose the uncert value lead the difference?

Thank you for your kindly help!