mrgsolve

mrgsolve facilitates simulation in R from hierarchical, ordinary differential equation (ODE) based models typically employed in drug development. See the example below.

Resources

Please see mrgsolve.github.io for additional resources.

Installation

We recommend staying up to date with the development version

devtools::install_github("metrumresearchgroup/mrgsolve")

Otherwise, install the latest release on CRAN

install.packages("mrgsolve", type="source")

Please be sure to see important install-related information here.

Ask a question

• Issue tracker (preferred) (requires GitHub account; ok for questions or issue reports)
• Google Group (email list or web-based discussion)

Example

library(mrgsolve)
library(dplyr)
library(ggplot2)

The model specification file is similar to other non-linear mixed effects modeling software

code <- '
$GLOBAL
#define CP (CENT/VC)
#define INH (CP/(IC50+CP))

$SET delta=0.1

$PARAM TVCL=1, TVVC=20, KA = 1.3, KIN=100, KOUT=2, IC50=10

$CMT GUT CENT RESP

$MAIN
double CL = exp(log(TVCL) + ETA(1));
double VC = exp(log(TVVC) + ETA(2));

RESP_0 = KIN/KOUT;

$OMEGA 0 0

$ODE
dxdt_GUT = -KA*GUT;
dxdt_CENT = KA*GUT - (CL/VC)*CENT;
dxdt_RESP = KIN*(1-INH) - KOUT*RESP;

$CAPTURE CP
'

The model is parsed, compiled, and dynamically loaded into the R session

• Information about the model is saved as an R object
• Important model attributes can be updated in R without recompiling

mod <- mcode("demo", code)

Use mrgsolve as an interactive simulation tool for model exploration and sensitivity analyses

• Simulated data are returned as R objects
• Input and output data are kept in memory in the R process; writing or reading to disk is never necessary (unless results are to be saved for later use).

out <- 
  mod %>%
  ev(amt=100, ii=24, addl=2) %>%
  mrgsim(end=120)

out

. Model:  demo.cpp 
. Dim:    1202 x 6 
. Time:   0 to 120 
. ID:     1 
.      ID time    GUT  CENT  RESP    CP
. [1,]  1  0.0   0.00  0.00 50.00 0.000
. [2,]  1  0.0 100.00  0.00 50.00 0.000
. [3,]  1  0.1  87.81 12.16 49.72 0.608
. [4,]  1  0.2  77.11 22.78 49.03 1.139
. [5,]  1  0.3  67.71 32.04 48.11 1.602
. [6,]  1  0.4  59.45 40.11 47.06 2.006
. [7,]  1  0.5  52.20 47.14 45.96 2.357
. [8,]  1  0.6  45.84 53.25 44.87 2.663

plot(out, CP+RESP~.)

out <- 
  mod %>% update(end=48) %>%
  ev(amt=100) %>%
  Req(CP,RESP) %>%
  knobs(TVVC=c(10,20,40), TVCL=c(0.5,1.5))

plot(out)

Use mrgsolve for large-scale population simulation

mod <- mod %>% omat(cmat(0.1, 0.67, 0.4))

Flexibility with input data sets

• Data set format that is likely familiar to modeling and simulation scientists
• No need to include observation records; mrgsolve will automatically insert

.data <- 
  expand.ev(ID=1:10, amt=c(100,300,1000)) %>%
  mutate(dose=amt)

head(.data)

.   ID amt evid cmt time dose
. 1  1 100    1   1    0  100
. 2  2 100    1   1    0  100
. 3  3 100    1   1    0  100
. 4  4 100    1   1    0  100
. 5  5 100    1   1    0  100
. 6  6 100    1   1    0  100

Input data are passed in as R objects

• Pass many different data sets or implement different designs in the same model code without recompiling
• Control simulation output from R to better manage memory

out <- 
  mod %>%
  data_set(.data) %>%
  Req(RESP,CP) %>% obsonly %>%
  carry_out(dose) %>%
  mrgsim(end=48, seed=1010)

plot(out, RESP~time|factor(dose), scales="same")

Pass simulated output to your favorite data summary or visualization routines

Summarise with dplyr

out %>%
  as.tbl %>%
  group_by(dose) %>%
  summarise(rmin = min(RESP), tmim=time[which.min(RESP)])

. # A tibble: 3 × 3
.    dose      rmin  tmim
.   <dbl>     <dbl> <dbl>
. 1   100 18.958869   2.9
. 2   300 16.117261   3.5
. 3  1000  6.198648   3.5

Plot with ggplot2

out %>%
  as.tbl %>%
  ggplot(data=.) +
  geom_line(aes(x=time, y=RESP, group=ID, col=factor(dose)))