Jump to content


Photo

Transit Comp Modelling using GUI in NLME

Transit comp NLME Graphical interface Ktr

  • Please log in to reply
14 replies to this topic

#1 harishkk

harishkk

    Advanced Member

  • Members
  • PipPipPip
  • 47 posts

Posted 27 July 2015 - 10:42 AM

Hello NLME modellers/Users,

Iam trying to model microspheres formulation of a drug using Transit Compartment modelling.I have used NLME Graphical Model building tool to build the Model.I have few issues for which i hope i will get solutions here

 

  1. I don't need Ktr (Transit rate constant) to be estimated as fixed effect 5 times, i need only 1 Ktr as fixed effect parameter (but while building the model NLME takes these as 5 Ktr parameters, due to which there is an error.So, Issue is "How do i tell NLME to consider 5Ktr as all equal and calculate only 1 Ktr instead of 5 Ktr
  2. Iam choosing the NLME way , as no other models are suitable for describing the data and i have sparse data in few subjects
  3. Iam attaching the PHX file for NLME user expert community to answer me

I hope i would get answer earlier 

Regards

Kaushik

Attached Files



#2 serge guzy

serge guzy

    Advanced Member

  • Members
  • PipPipPip
  • 485 posts

Posted 27 July 2015 - 02:15 PM

Hello NLME modellers/Users,

Iam trying to model microspheres formulation of a drug using Transit Compartment modelling.I have used NLME Graphical Model building tool to build the Model.I have few issues for which i hope i will get solutions here

 

  1. I don't need Ktr (Transit rate constant) to be estimated as fixed effect 5 times, i need only 1 Ktr as fixed effect parameter (but while building the model NLME takes these as 5 Ktr parameters, due to which there is an error.So, Issue is "How do i tell NLME to consider 5Ktr as all equal and calculate only 1 Ktr instead of 5 Ktr
  2. Iam choosing the NLME way , as no other models are suitable for describing the data and i have sparse data in few subjects
  3. Iam attaching the PHX file for NLME user expert community to answer me

I hope i would get answer earlier 

Regards

Kaushik

Dear Kaushik

I made a model for you to answer to your issues.it is attached.

1: you debone as model parameter ktr and link it physically to all ktr parameters showing up in your flow between transit compts

2: you cannot use a number for compartments.  change 1,2,3.. into AT1,AT2,..

3: I used NCA to initialize V,Cl,V2 and Q.

4: I used 10 x lambdaz as initial value for Ka1 and Ka2

5: I did the base model

6: I assumed (you can change that) that potentially Cl,V Ka1 and Ka2 can be linked to source and/or formulation

7: I added 2 columns with integer values for formulation and source(species)

8: I used covariate relationships rather than sorting to see what are the potential covariate

9: I used covariate search after finishing the base model

10: the final model is the last model and show Cl with source and V with formulation as only statistically significant covariate relationships.

I hope it helps you to continue your investigation.

best Regards

Serge

Attached Files



#3 harishkk

harishkk

    Advanced Member

  • Members
  • PipPipPip
  • 47 posts

Posted 11 September 2015 - 07:15 AM

Hello Sergy/members

I am trying to build the below transit model as per the paper attached using ASCII model option in WNL,i have tried several times but i have a problem in fitting

Can anyone help me in identifying the problem

I know PHX Modelling language could be used but, i am more comfortable with using ASCII at present

I have attached the Phx project file along with the ASCII code, when i try to run this ascii Code it took > 4 hrs for me to execute and finally give answer failed to fit, Iam also attaching the paper based on which this ascii Model is written

 

Eagerly waiting for the response

 

Regards

Harish

Attached Files



#4 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,316 posts

Posted 11 September 2015 - 08:49 AM

Harish, the project you attached does not contain the model/failed run results - if you could attach that then it might help us debug your issue.

 

 However we have strictly limited support for the old ascii language as it is essentially obsolete and any bugs or other problems with it will not be addressed since it was replaced by the Phoenix modelling language back in 2009.  I therefore higly recommend you try again with PML, there are several transit compartment examples already posted here and elsewhere.

 

   Simon.



#5 serge guzy

serge guzy

    Advanced Member

  • Members
  • PipPipPip
  • 485 posts

Posted 11 September 2015 - 09:23 AM

Yes please, explain what you are trying to do and I can help you with the pml code.

You need to shift to the pml as it is more versatile, enable single, multiple doses with transit compartments, has full support system to guide you. The old winnonlion ACI code is to me gone and soon or later it will not be used at all.

Therefore, small help and you can do it with pml but I need to know what you are trying to model.

best

Serge



#6 harishkk

harishkk

    Advanced Member

  • Members
  • PipPipPip
  • 47 posts

Posted 14 September 2015 - 09:19 AM

Hello Serge,

Iam interested in modelling the invivo PCT of Diclofenac Microsphere formulations, i have tried to use transit compartment for modelling the 2 peaks observed due to initial burst release and later sustained release

you can look at the XY plot which clearly demonstrates the double peaks observed and now my question is trying transit comp model iam trying to fit the model to the data but in vain, if you can help me do it , it would be of great help

 

Regards

Harish

 

Note: Papers i have already attached earlier

Attached Files



#7 serge guzy

serge guzy

    Advanced Member

  • Members
  • PipPipPip
  • 485 posts

Posted 14 September 2015 - 11:23 AM

Hello Serge,

Iam interested in modelling the invivo PCT of Diclofenac Microsphere formulations, i have tried to use transit compartment for modelling the 2 peaks observed due to initial burst release and later sustained release

you can look at the XY plot which clearly demonstrates the double peaks observed and now my question is trying transit comp model iam trying to fit the model to the data but in vain, if you can help me do it , it would be of great help

 

Regards

Harish

 

Note: Papers i have already attached earlier

Dear Harish

I strongly suggest you to take an introductory course on PK/PD modeling or introduction to Population PK/PD with Phoenix NLME.

Here you have a model with 2 absorption sites. Putting 2 absorptions with fraction associated with each site and a tlag enables to get a very good fit. The project is attached.

The model building in Phoenix requires a course as there is a lot to learn I would like you to got through.

There is not need to use transit compartment as you can see from the fitting properties.

 

The final parameters cannot give you good standard error properties but 2 compartment is way better than 1. The best would be to get the individual data and NOT the mean and perform a population approach. This will enable you to get better statistical properties.

 

tvV 0.0183846
tvCl 0.00199638
tvKa1 65.6666
tvKa2 109.519
tvf 1.12077
tvtlag 3.99833
tvV2 0.136579
tvQ 0.000184485
stdev0 0.096293
 

Attached Files



#8 harishkk

harishkk

    Advanced Member

  • Members
  • PipPipPip
  • 47 posts

Posted 15 September 2015 - 11:57 AM

Hello Serge,

Thanks for sharing the phx proj file, i want to simulate at these time points using the model_2c which you have shared with me earlier can you help with these

 

Regards

0 1 2 4 6 8 12 24 192 360 528 696 697 698 700 702 704 708 720

#9 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,316 posts

Posted 15 September 2015 - 01:31 PM

Harish, as Serge said you would likely benefit a lot from some training or at least working through the examples in the WNL and Phoenix NLME 1.3 User's Guide.pdf.

 

  In the attached project I simply copied the last model and ran in Sim mode requestiung the time points you asked for but you could do this more 'intelligently' with e.g. Seq statements etc.  look at the Simulation Table and compare with the Table01

 

  SimonAttached File  ms_serge_add_sim.phxproj   1.64MB   456 downloads

Attached Thumbnails

  • ind_simulation.jpg


#10 harishkk

harishkk

    Advanced Member

  • Members
  • PipPipPip
  • 47 posts

Posted 16 September 2015 - 09:42 AM

Thanks Simon,

I am learning the tricks of moving to PML from WNL ASCII models and slowly increasing my knowledge on building models in the new PML / GUI,

In continuation to my earlier response, iam keen to build simultaneous plasma and skin concnetration model fitting for which i have built the model (image  and resource paper attached) iam also attaching the PHX project file  with the model built, iam confused as to how the input data should be for simultaneous fitting

i have provided both the plasma and skin conc data in the project 

 

your help will be of great support to me

 

Regards

Harish

Attached Files



#11 serge guzy

serge guzy

    Advanced Member

  • Members
  • PipPipPip
  • 485 posts

Posted 16 September 2015 - 12:18 PM

Thanks Simon,

I am learning the tricks of moving to PML from WNL ASCII models and slowly increasing my knowledge on building models in the new PML / GUI,

In continuation to my earlier response, iam keen to build simultaneous plasma and skin concnetration model fitting for which i have built the model (image  and resource paper attached) iam also attaching the PHX project file  with the model built, iam confused as to how the input data should be for simultaneous fitting

i have provided both the plasma and skin conc data in the project 

 

your help will be of great support to me

 

Regards

Harish

Dear Harish

You can use the graphical mode to define the model as in the paper because the Plasma PK model and the skin model do not follow strict mass balance assumption. You need to shift to edit as textual and write the code based on the differential equations as in the paper. It will take me some time but I can do it if you want. Now is your plasma model still with double peak?

the steps should be

1: do the Plasma model and you can use graphical as I did

2: do a template graphical model for the skin

3: go to textual and remove the follow from plasma to skin that shows up in the plasma mass balance (that is the assumption of the model).

This is quite complex and I will need to explain how to do all that but lt me know if this is what you want.

Best

Serge



#12 serge guzy

serge guzy

    Advanced Member

  • Members
  • PipPipPip
  • 485 posts

Posted 16 September 2015 - 12:36 PM

If you do that, then you will get the following code

 

test(){
# deriv(A1 = - (Cl * C) + (Aa1 * Ka1) + (Aa2 * Ka2)- (Q * (C - C2))- (Kin * C))

 deriv(A1 = - (Cl * C) + (Aa1 * Ka1) + (Aa2 * Ka2)- (Q * (C - C2)))

 urinecpt(A0 = (Cl * C))
 deriv(Aa1 = - (Aa1 * Ka1))
 deriv(Aa2 = - (Aa2 * Ka2))
 deriv(A2 = (Q * (C - C2)))
 deriv(A1s = (Kin * C)- (Kout * CS1)- (KS12 * CS1) + (A2S * KS21))
 urinecpt(A0skin = (Kout * CS1))
 deriv(A2S = (KS12 * CS1)- (A2S * KS21))
 C = A1 / V
 dosepoint(Aa1, bioavail = (ilogit(f)), idosevar = Aa1Dose, infdosevar = Aa1InfDose, infratevar = Aa1InfRate)
 error(CEps = 0.096293)
 observe(CObs = C * (1 + CEps))
 dosepoint(Aa2, tlag = (tlag), bioavail = (1-ilogit(f)), idosevar = Aa2Dose, infdosevar = Aa2InfDose, infratevar = Aa2InfRate)
 C2 = A2 / V2
 CS1 = A1s / Vs
 C2 = A2S / V2SS
 error(CEpsCS1 = 0.1)
 observe(CObsSKIN = CS1 * (1 + CEpsCS1))
 stparm(V = tvV)
 stparm(Cl = tvCl)
 stparm(Ka1 = tvKa1)
 stparm(Ka2 = tvKa2)
 stparm(f = tvf)
 stparm(tlag = tvtlag)
 stparm(V2 = tvV2)
 stparm(Q = tvQ)
 stparm(Vs = tvVs)
 stparm(Kin = tvKin)
 stparm(Kout = tvKout)
 stparm(V2SS = tvV2SS)
 stparm(KS12 = tvKS12)
 stparm(KS21 = tvKS21)
 fixef(tvV = c(0, 0.0183846, ))
 fixef(tvCl = c(0, 0.00199638, ))
 fixef(tvKa1 = c(0, 65.6666, ))
 fixef(tvKa2 = c(0, 109.519, ))
 fixef(tvf = c(, 1.12077, ))
 fixef(tvtlag = c(0, 3.99833, ))
 fixef(tvV2 = c(, 0.136579, ))
 fixef(tvQ = c(, 0.000184485, ))
 fixef(tvVs = c(, 1, ))
 fixef(tvKin = c(, 1, ))
 fixef(tvKout = c(, 1, ))
 fixef(tvV2SS = c(, 1, ))
 fixef(tvKS12 = c(, 1, ))
 fixef(tvKS21 = c(, 1, ))
 secondary(fabsolute = ilogit(tvf))
}

 



#13 serge guzy

serge guzy

    Advanced Member

  • Members
  • PipPipPip
  • 485 posts

Posted 16 September 2015 - 12:38 PM

Here is the model based on the paper but PK plasma part as before with double peak.

You can use this as a template. Put the skin data in the same data set as the plasma. Just use a different column for skin than Plasma. You can use my template and add the data.

Hope it helps/.

Best

Serge

Attached Files



#14 harishkk

harishkk

    Advanced Member

  • Members
  • PipPipPip
  • 47 posts

Posted 06 October 2015 - 05:11 AM

Here is the model based on the paper but PK plasma part as before with double peak.

You can use this as a template. Put the skin data in the same data set as the plasma. Just use a different column for skin than Plasma. You can use my template and add the data.

Hope it helps/.

Best

Serge

Thanks Sergy,

The skin conc are at different timepoints and plasma conc are at diff time points so adding the skin conc column beside plasma conc column, would not be correct, can we denote it as function 1  and 2

 

please let me know

 

Regards

Harish



#15 serge guzy

serge guzy

    Advanced Member

  • Members
  • PipPipPip
  • 485 posts

Posted 06 October 2015 - 10:49 AM

Dear Harish

It does not matter. You just use one row with let say the plasma level showing up in the plasma column and in the time column you put the time at which the plasma level is measured.When you have a skin level you use another with the time at which the skin level was measured and the level in the skin column.

For example suppose a dose in the plasma compt (A1) of 100, an observation in plasma at t=2 and a skin observation at t=4. Here is the template data set for one hypothetical patient.

 

id     time skin plasma A1   

1       0                         100

1       2                 87

1      4       43 

 

 

Same for each patient

 

Let me know if you need more explanation. I am out of town but can access my email every evening.

Best

Serge







Also tagged with one or more of these keywords: Transit comp, NLME, Graphical interface, Ktr

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users