Closed cmneil closed 5 years ago
Dear Christopher:
Thanks for your interest in nlmixr.
What the error msg means is that although the optimization step is successful, the covariance step (using optimHess) is not. Have your tried the dynmodel's cousin dynmodel.mcmc? The syntax is almost the same, but it would spare you the covariance step. You could then summarize the mcmc samples for s.e., etc.
Best, Wenping
Dear Christopher:
Thanks for your interest in nlmixr.
What the error msg means is that although the optimization step is successful, the covariance step (using optimHess) is not. Have your tried the dynmodel's cousin dynmodel.mcmc? The syntax is almost the same, but it would spare you the covariance step. You could then summarize the mcmc samples for s.e., etc.
Best, Wenping
On Wed, Mar 7, 2018 at 12:51 PM, cmneil notifications@github.com wrote:
Hello, I'm trying to use this package for modeling non-population data for a single-compartment model. I thought dynmodel might be the function that could help me to that end, but I keep getting an optimHess error when running it.
Does anyone have any guidance on how to use this package for non-population data, or using single-compartment models on the average of non-population data?
Best, Christopher Neil
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/nlmixrdevelopment/nlmixr/issues/33, or mute the thread https://github.com/notifications/unsubscribe-auth/AMdZdnZoNzo0EqMDplr5kpW6nMmP8D41ks5tcB4dgaJpZM4Sg4q- .
Dear Wenping,
I am very interested in the possibilities of nlmixr.
I am a little confused, because I am attempting to use these functions to model the very simplest system, but they do not seem to be working. I can get the example data to function for its example system, but not any other system. I believe the syntax of my system is correct, and the functions try to solve them, but the results are not correct. The results are not better with the mcmc version either. In that case, a plotting error arises over x and y not being the same length.
Also, I have used RxODE to simulate data that are close to my observations. This indicates to me that the issue is not in the system of odes, but perhaps in the modeling algorithm. Is it possible that you could take a look at the code below and see if there is a reason for the issue? I hope it is not too much to ask. I appreciate the time.
Best, Christopher Neil
ode <- "
C2 = centr/V2;
d/dt(depot) =-KA*depot;
d/dt(centr) = KA*depot - CL*(centr/V2);
"
sys2 <- RxODE(model = ode)
mod <- cp ~ C2 + prop(.1)
inits2 <- c(KA = 3, V2 = 0.02, CL = 0.0117)
fixPars2 <- NULL
ev2 <- eventTable()
ev2$add.sampling(dat2$time)
ev2$add.dosing(dose = 2.5, amount.units = "umol")
(fit <- dynmodel(sys2, mod, ev2, inits2, dat2, fixPars2)))
dat2
time cp
2 2 25.400000
3 8 0.790320
4 10 0.232750
5 12 0.137750
6 16 0.007320
7 24 0.014625
On Wed, Mar 7, 2018 at 3:19 PM, wwang-at-github notifications@github.com wrote:
Dear Christopher:
Thanks for your interest in nlmixr.
What the error msg means is that although the optimization step is successful, the covariance step (using optimHess) is not. Have your tried the dynmodel's cousin dynmodel.mcmc? The syntax is almost the same, but it would spare you the covariance step. You could then summarize the mcmc samples for s.e., etc.
Best, Wenping
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/nlmixrdevelopment/nlmixr/issues/33#issuecomment-371270202, or mute the thread https://github.com/notifications/unsubscribe-auth/Afz-yfVunCr3XGD7yKexbI85veVB-RIVks5tcEAHgaJpZM4Sg4q- .
--
Christopher M. Neil, M.A.
Graduate Scientist
AstraZeneca Pharmaceuticals LP
R&D Boston | Discovery Sciences | Mechanistic Biology & Profiling
(+1)720-254-7509
neil.christopher.michael@gmail.com
Thanks Christopher for your interest in nlmixr. I'll take a look & let you know my take.
Incidentally, what's the units of your dose, concentration & your pars? What's the true value you used in simulation?
Best, Wenping
On Thu, Mar 8, 2018 at 3:35 PM, cmneil notifications@github.com wrote:
Dear Wenping,
I am very interested in the possibilities of nlmixr.
I am a little confused, because I am attempting to use these functions to model the very simplest system, but they do not seem to be working. I can get the example data to function for its example system, but not any other system. I believe the syntax of my system is correct, and the functions try to solve them, but the results are not correct. The results are not better with the mcmc version either. In that case, a plotting error arises over x and y not being the same length.
Also, I have used RxODE to simulate data that are close to my observations. This indicates to me that the issue is not in the system of odes, but perhaps in the modeling algorithm. Is it possible that you could take a look at the code below and see if there is a reason for the issue? I hope it is not too much to ask. I appreciate the time.
Best, Christopher Neil
ode <- " C2 = centr/V2; d/dt(depot) =-KA*depot; d/dt(centr) = KA*depot - CL*(centr/V2); " sys2 <- RxODE(model = ode) mod <- cp ~ C2 + prop(.1) inits2 <- c(KA = 3, V2 = 0.02, CL = 0.0117) fixPars2 <- NULL ev2 <- eventTable() ev2$add.sampling(dat2$time) ev2$add.dosing(dose = 2.5, amount.units = "umol") (fit <- dynmodel(sys2, mod, ev2, inits2, dat2, fixPars2))) dat2 time cp 2 2 25.400000 3 8 0.790320 4 10 0.232750 5 12 0.137750 6 16 0.007320 7 24 0.014625
On Wed, Mar 7, 2018 at 3:19 PM, wwang-at-github notifications@github.com wrote:
Dear Christopher:
Thanks for your interest in nlmixr.
What the error msg means is that although the optimization step is successful, the covariance step (using optimHess) is not. Have your tried the dynmodel's cousin dynmodel.mcmc? The syntax is almost the same, but it would spare you the covariance step. You could then summarize the mcmc samples for s.e., etc.
Best, Wenping
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/nlmixrdevelopment/nlmixr/issues/33#issuecomment- 371270202, or mute the thread https://github.com/notifications/unsubscribe-auth/Afz- yfVunCr3XGD7yKexbI85veVB-RIVks5tcEAHgaJpZM4Sg4q- .
--
Christopher M. Neil, M.A.
Graduate Scientist
AstraZeneca Pharmaceuticals LP
R&D Boston | Discovery Sciences | Mechanistic Biology & Profiling
(+1)720-254-7509 <(720)%20254-7509>
neil.christopher.michael@gmail.com
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/nlmixrdevelopment/nlmixr/issues/33#issuecomment-371615848, or mute the thread https://github.com/notifications/unsubscribe-auth/AMdZdlvDYGn4u6e5atKKaQBKIIYP79UGks5tcZYSgaJpZM4Sg4q- .
Hi Wenping,
The dose was 2.5umol, the concentrations are in uM, the pars were calculated to be around 0.0117L/h for CL, 0.02L for V2, and 3.4h^-1 for KA. These were calculated in Phoenix and successfully simulated with RxODE.
Hope that’s what you were looking for.
Thank you!
Sent from my iPhone
On Mar 9, 2018, at 13:39, wwang-at-github notifications@github.com wrote:
Thanks Christopher for your interest in nlmixr. I'll take a look & let you know my take.
Incidentally, what's the units of your dose, concentration & your pars? What's the true value you used in simulation?
Best, Wenping
On Thu, Mar 8, 2018 at 3:35 PM, cmneil notifications@github.com wrote:
Dear Wenping,
I am very interested in the possibilities of nlmixr.
I am a little confused, because I am attempting to use these functions to model the very simplest system, but they do not seem to be working. I can get the example data to function for its example system, but not any other system. I believe the syntax of my system is correct, and the functions try to solve them, but the results are not correct. The results are not better with the mcmc version either. In that case, a plotting error arises over x and y not being the same length.
Also, I have used RxODE to simulate data that are close to my observations. This indicates to me that the issue is not in the system of odes, but perhaps in the modeling algorithm. Is it possible that you could take a look at the code below and see if there is a reason for the issue? I hope it is not too much to ask. I appreciate the time.
Best, Christopher Neil
ode <- " C2 = centr/V2; d/dt(depot) =-KA*depot; d/dt(centr) = KA*depot - CL*(centr/V2); " sys2 <- RxODE(model = ode) mod <- cp ~ C2 + prop(.1) inits2 <- c(KA = 3, V2 = 0.02, CL = 0.0117) fixPars2 <- NULL ev2 <- eventTable() ev2$add.sampling(dat2$time) ev2$add.dosing(dose = 2.5, amount.units = "umol") (fit <- dynmodel(sys2, mod, ev2, inits2, dat2, fixPars2))) dat2 time cp 2 2 25.400000 3 8 0.790320 4 10 0.232750 5 12 0.137750 6 16 0.007320 7 24 0.014625
On Wed, Mar 7, 2018 at 3:19 PM, wwang-at-github notifications@github.com wrote:
Dear Christopher:
Thanks for your interest in nlmixr.
What the error msg means is that although the optimization step is successful, the covariance step (using optimHess) is not. Have your tried the dynmodel's cousin dynmodel.mcmc? The syntax is almost the same, but it would spare you the covariance step. You could then summarize the mcmc samples for s.e., etc.
Best, Wenping
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/nlmixrdevelopment/nlmixr/issues/33#issuecomment- 371270202, or mute the thread https://github.com/notifications/unsubscribe-auth/Afz- yfVunCr3XGD7yKexbI85veVB-RIVks5tcEAHgaJpZM4Sg4q- .
--
Christopher M. Neil, M.A.
Graduate Scientist
AstraZeneca Pharmaceuticals LP
R&D Boston | Discovery Sciences | Mechanistic Biology & Profiling
(+1)720-254-7509 <(720)%20254-7509>
neil.christopher.michael@gmail.com
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/nlmixrdevelopment/nlmixr/issues/33#issuecomment-371615848, or mute the thread https://github.com/notifications/unsubscribe-auth/AMdZdlvDYGn4u6e5atKKaQBKIIYP79UGks5tcZYSgaJpZM4Sg4q- .
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.
Hi Wenping,
I was just curious if there had been any development to determine what the issue might be.
Best, Christopher Neil
Hi Christopher,
Would you mind uploading the data file, or the RxODE script you used for simulations? That would be very helpful.
Hi Justin,
I was following up with Wenping from about a month ago. Everything should be available from earlier in this thread.
Best, Chris
On Fri, Apr 6, 2018 at 9:07 AM, Justin Wilkins notifications@github.com wrote:
Hi Christopher,
Would you mind uploading the data file, or the RxODE script you used for simulations? That would be very helpful.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/nlmixrdevelopment/nlmixr/issues/33#issuecomment-379247608, or mute the thread https://github.com/notifications/unsubscribe-auth/Afz-ye3xjfaABCrF4i9QF2NMCiHoK1Rmks5tl2iRgaJpZM4Sg4q- .
--
Christopher M. Neil, M.A.
Graduate Scientist
AstraZeneca Pharmaceuticals LP
R&D Boston | Discovery Sciences | Mechanistic Biology & Profiling
(+1)720-254-7509
neil.christopher.michael@gmail.com
Hi Chris,
With current GitHub versions of nlmixr and RxODE, both of the following statements deliver results:
fit <- dynmodel(sys2, mod, ev2, inits2, dat2, fixPars2)
fit2 <- dynmodel.mcmc(sys2, mod, ev2, inits2, dat2, fixPars2) # very slow with default number of iterations
Answers are, however, wildly divergent. This may be because the input data is very limited (6 records) with very little information in the absorption phase. Have you tried with a larger dataset?
Hi Justin,
I have not tried a larger dataset. However, my work is pre-clinical and it usually has to work with a small number of time points based on in vivo studies. I don't have data from the absorption phase. These are my results from inputting those two statements:
ode <- "
- C2 = centr/V2;
- d/dt(depot) =-KA*depot;
- d/dt(centr) = KAdepot - CL(centr/V2);
- "
sys2 <- RxODE(model = ode) (fit2 <- dynmodel(sys2, mod, ev2, inits2, dat2, fixPars2)) Error in optimHess(fit$par, obj) : non-finite finite-difference value [1] (fit2 <- dynmodel.mcmc(sys2, mod, ev2, inits2, dat2, fixPars2)) mean sd cv% KA 31.852878 31.975560 100.38515 V2 3.119708 4.045197 129.66587 CL 265.669469 168.504815 63.42649 err1 24.582880 31.752407 129.16472
plot(fit2) Error in xy.coords(x, y, xlabel, ylabel, log) : 'x' and 'y' lengths differ
I did get results from dynmodel.mcmc, but it seems to be not useful. I'm not sure of what to make of the error from plotting it.
Best, Chris
On Fri, Apr 6, 2018 at 10:23 AM, Justin Wilkins notifications@github.com wrote:
Hi Chris,
With current GitHub versions of nlmixr and RxODE, both of the following statements deliver results:
fit <- dynmodel(sys2, mod, ev2, inits2, dat2, fixPars2) fit2 <- dynmodel.mcmc(sys2, mod, ev2, inits2, dat2, fixPars2) # very slow with default number of iterations
Answers are, however, wildly divergent. This may be because the input data is very limited (6 records) with very little information in the absorption phase. Have you tried with a larger dataset?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/nlmixrdevelopment/nlmixr/issues/33#issuecomment-379269117, or mute the thread https://github.com/notifications/unsubscribe-auth/Afz-yeZIvDXfisKtc5k__fMUGFRKj_JDks5tl3pvgaJpZM4Sg4q- .
--
Christopher M. Neil, M.A.
Graduate Scientist
AstraZeneca Pharmaceuticals LP
R&D Boston | IMED Oncology | Drug Metabolism & Pharmacokinetics
(+1)720-254-7509
neil.christopher.michael@gmail.com
I don't get any errors when running dynmodel
with the versions of RxODE and nlmixr I installed this afternoon, but the answers are very odd, and plotting does not work. We'll continue looking into it.
Thank you. And the version should be up-to-date. I just reinstalled on a new laptop last Monday.
Best, Chris
On Fri, Apr 6, 2018 at 11:10 AM, Justin Wilkins notifications@github.com wrote:
I don't get any errors when running dynmodel with the versions of RxODE and nlmixr I installed this afternoon, but the answers are very odd, and plotting does not work. We'll continue looking into it.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/nlmixrdevelopment/nlmixr/issues/33#issuecomment-379283656, or mute the thread https://github.com/notifications/unsubscribe-auth/Afz-yeZyf1Twpzm_yQM1qj3I_NR0D7vdks5tl4VjgaJpZM4Sg4q- .
--
Christopher M. Neil, M.A.
Graduate Scientist
AstraZeneca Pharmaceuticals LP
R&D Boston | IMED Oncology | Drug Metabolism & Pharmacokinetics
(+1)720-254-7509
neil.christopher.michael@gmail.com
I think this is resolved @cmneil . If not, please let me know.
Hello, I'm trying to use this package for modeling non-population data for a single-compartment model. I thought dynmodel might be the function that could help me to that end, but I keep getting an optimHess error when running it.
Does anyone have any guidance on how to use this package for non-population data, or using single-compartment models on the average of non-population data?
Best, Christopher Neil