19 replies to this topic

[Aiqun Li]

Hi there, I have a naive question regarding the Michaelis-Menten modeling by phoenix and WNL classic modeling. As show in the attached phoenix project. Estimated values are so different: WNL Classic _MM V 33.31669 VM 10000.05 KM 2999.982 Phoenix_MM tvV 41.1545 tvKm 2004510 tvVmax 1.36E+08 Another question is how to estimate the initial values for the WNL classic MM modeling? Thank you very much for your help! ali [file name=MM_example_301.phxproj size=545809]http://www.pharsight.com/extranet/media/kunena/attachments/legacy/files/MM_example_301.phxproj[/file]

Attached Files

•  MM_example_301.phxproj   533.02KB   366 downloads

Edited by Simon Davis, 16 January 2018 - 10:36 AM.

[Simon Davis]

Ali,
Look at the WNL classic fit; it is not very good right?


so take a look at the WARNINGS AND ERRORS tab

Modeling error 3
*** WARNING 10134 *** PARAMETER ESTIMATE FOR KM IS AT OR NEAR BOUNDARY.

Core Output may have more information

So the clue is you have restricted the search space too much for WNL classic to find the optimal solution.

typically I would suggest and upper boundary of 10 times your estimate e.g. 450000 for vmax and also you will need to increase Km and you will find that you now get a better fit with very similar estimates for Volume.

there are still differences for the other parameters but generally you will find the Phoenix algorithm more flexible and powerful, the WNL engine is mainly there for backwards compatibility.

I would also check my units carefully as these large numbers suggest to me you may not be working in the optimum order of magnitude

Simon

Attached Thumbnails

• 

[Aiqun Li]

Dear Simon: Thank you very much for the expllanation. I agree with you. Now I have a related question. In the attached phoenix project, I run the phoeninx MM modeling first and got the parameters, then I put the same parameters as initial estimates into the WNL classic MM model. Surprisely the fiting is very poor. I am very puzzling. What is wrong with my fiting? Your kind help is highly appreciated! With best regards, ali [file name=Rat__IV_Bolus_Plasma.phxproj size=570742]http://www.pharsight.com/extranet/media/kunena/attachments/legacy/files/Rat__IV_Bolus_Plasma.phxproj[/file]

Attached Files

•  Rat__IV_Bolus_Plasma.phxproj   557.37KB   351 downloads

Edited by Simon Davis, 16 January 2018 - 10:37 AM.

[Samer Mouksassi]

Dear Ali,

Fitting MM models can be difficult with limited data.

In your phoenix model we see that we miss the last concentration value at time =2

look at the plot ind dv, ipred vs tad with log y axis

what was the limit of the quantitation of your assay ?

 

for the winonlin calssic model go to engine setting and select nelder mead

 

I got

 

Parameter Estimate

V 23.388996

VM 345398.84

KM 48258.297

 

 

with the intials used

 

Parameter Estimate

V 20

VM 8000000

KM 48000

 

fits were similar between the two

[serge guzy]

Dear Ali

First of all, in Phoenix modeling (without checking the population box), the equation for the 1 compartment model is described by

 

deriv(A1 = - Vmax * C / (Km + C))

 

A1 is amount and therefore Vmax has the units of amount per unit of time, Km has units of concentration

 

In winnonlin classic, based on the picture, I can see that the equation is

 

dC/dt=-Vmax*C/(Km+C)

 

Therefore here Vmax has units of concentration per unit of time and Km has concentration units again.

 

Therefore Vmax does not have same units in the two models.

 

 

In addition, the optimization method is different. Phoenix Model uses maximum likelihood while the classic Winnolnlin uses least square.

 

Now if you start with initial estimates for Vmax that apply only for one of the two models, these can be way off for the other model and it can explain why you could have numerical difficulties.

This in addition of the relevant responses Simon gave you.

 

If you want to compare, use the Vmax of winnonlin classic and x by the V to get the Vmax of Phoenix model and try again.

Again all this is what I see with the winnonlin picture (equation shows up below the graph).

I am really not an expert in Winnonlin like Simon.

 

Best

Serge

[serge guzy]

Dear Ali

Look at your lower and upper bounds you choose for the winnonlin classic model. You have one parameter that have initial less than lower.

I do not know how the classic winnonlin uses that but it is not right.

Then adjust the initial for Vmax to reflect difference in units and changes the lower and upper to have initial in between and then compare.

best

Serge

[serge guzy]

Dear Ali I solved the problem. The project is attached. If you take final estimates of Vmax from Phoenix (amount per unit of time) and divide by Volume, you will get the Vmax showing up in winnonlin classical. The results match almost perfectly when you use the same boundaries (adjusted for Vmax) for the model parameters. Does that help? Best [file name=MM_example_official_serge.phxproj size=560467]http://www.pharsight.com/extranet/media/kunena/attachments/legacy/files/MM_example_official_serge.phxproj[/file]

Attached Files

•  MM_example_official_serge.phxproj   547.33KB   345 downloads

Edited by Simon Davis, 16 January 2018 - 10:37 AM.

[Aiqun Li]

Dear Simon and Serge:

 

Thank you so much for the response. I really appreciate it.

 

Have a blessed holiday season and Happy New Year!

 

With best regards,

 

Aiqun

[Aiqun Li]

Hi Serge:

 

Thanks again for answering my question. For some reason, I can not open your attached file. It says I need a password.......

 

Best regards,

 

Aiqun

[serge guzy]

Dear Ali

Please see the link attached that should allow you to grab the project.

best Regards;

Serge

 

https://app.box.com/s/af52t053qqxm2a56tmyt

[Aiqun Li]

Serge:

 

Thank you for your help. It works for this one. But I have the same problem for the other posters you answered the questions and attached a phoenix project file, such as the transduction model.

 

Could you please help? Why does this happen? I do not want to bother you everytime.....

 

Best regards,

 

Aiqun

[serge guzy]

Can somebody help Aiqun with downloading projects. I think it can be related to the browser.

Can you try to just double click the project and see if it opens?

For the meantime can you send me the reference used for the transduction model.

I will then check where I put it.

best

Serge

[Aiqun Li]

Serge:

 

Thanks. The transduction model ref# is 2206. I tried double click and it came up with a zip file and asking for password. Bascially the same question.

 

Thank you for your help!

 

Aiqun

[Samer Mouksassi]

Hi,

As Serge points out this is a browser issue.

One workaround is that this zip file your downloaded change the extension from .zip to .phxproj and then you can double click and it opens in phoenix.

 

 

Samer

[Aiqun Li]

Samer and Serge:

 

Thank you so much! It works! Problem solved!! Much appreciated!!

 

One more question, could you provide a example for TMDD model? The reason I am asking this question is I want to compare MM Vs TMDD.

 

Thank you in advance!

 

Aiqun

[serge guzy]

Here is one TMDD simulated model. The model is fit to these simulated data ( to make sure there is no confidentiality issue here).

I hope it helps.

The link is shown below.

best Regards

Serge

 

https://app.box.com/s/pefw4qeyg81v9qdmhkyj

[Aiqun Li]

Serge:

 

Thank you so so much!!

 

Have a nice weekend!

 

Aiqun

[Simon Davis]

Aiqun,

    This has been discussed on the forum a few times and it’s an artefact of the browser typically IE, e.g.;

 

http://pharsight.com/extranet/index.php?option=com_kunena&Itemid=55&func=view&catid=51&id=1908#1919

 

if you're using IE in particular as your browser it "helpfully" renames the PHXPROJ file to Zip (since it is in fact a compressed file). Just download the file then rename the extension back to xxx.PHXPROJ, there is no password protection on Serge's file when opened in Phoenix.

 

   Simon.

[Simon Davis]

For TMDD examples you could also use the "search forum" function e.g. http://pharsight.com...51&id=2159#2162

[Aiqun Li]

Simon:

 

Thanks for your both replies. I do use IE and I will search more TMDD examples in the forum.

 

Best regards,

 

Aiqun