1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
|
\name{mkinfit}
\alias{mkinfit}
\title{
Fit a kinetic model to data with one or more state variables.
}
\description{
This function uses the Flexible Modelling Environment package
\code{\link{FME}} to create a function calculating the model cost, i.e. the
deviation between the kinetic model and the observed data. This model cost is
then minimised using the Levenberg-Marquardt algorithm \code{\link{nls.lm}},
using the specified initial or fixed parameters and starting values.
In each step of the optimsation, the kinetic model is solved using the
function \code{\link{mkinpredict}}. The variance of the residuals for each
observed variable can optionally be iteratively reweighted until convergence
using the argument \code{reweight.method = "obs"}.
}
\usage{
mkinfit(mkinmod, observed,
parms.ini = "auto",
state.ini = c(100, rep(0, length(mkinmod$diffs) - 1)),
fixed_parms = NULL, fixed_initials = names(mkinmod$diffs)[-1],
solution_type = "auto",
method.modFit = "Marq",
control.modFit = list(),
plot = FALSE, quiet = FALSE, err = NULL, weight = "none",
scaleVar = FALSE,
atol = 1e-8, rtol = 1e-10, n.outtimes = 100,
reweight.method = NULL,
reweight.tol = 1e-8, reweight.max.iter = 10,
trace_parms = FALSE, ...)
}
\arguments{
\item{mkinmod}{
A list of class \code{\link{mkinmod}}, containing the kinetic model to be fitted to the data.
}
\item{observed}{
The observed data. It has to be in the long format as described in
\code{\link{modFit}}, i.e. the first column called "name" must contain the
name of the observed variable for each data point. The second column must
contain the times of observation, named "time". The third column must be
named "value" and contain the observed values. Optionally, a further column
can contain weights for each data point. If it is not named "err", its name
must be passed as a further argument named \code{err} which is then passed
on to \code{\link{modFit}}.
}
\item{parms.ini}{
A named vector of initial values for the parameters, including parameters
to be optimised and potentially also fixed parameters as indicated by
\code{fixed_parms}. If set to "auto", initial values for rate constants
are set to default values. Using parameter names that are not in the model
gives an error.
It is possible to only specify a subset of the parameters that the model
needs. You can use the parameter lists "bparms.ode" from a previously
fitted model, which contains the differential equation parameters from this
model. This works nicely if the models are nested. An example is given
below.
}
\item{state.ini}{
A named vector of initial values for the state variables of the model. In
case the observed variables are represented by more than one model
variable, the names will differ from the names of the observed variables
(see \code{map} component of \code{\link{mkinmod}}). The default is to set
the initial value of the first model variable to 100 and all others to 0.
}
\item{fixed_parms}{
The names of parameters that should not be optimised but rather kept at the
values specified in \code{parms.ini}.
}
\item{fixed_initials}{
The names of model variables for which the initial state at time 0 should
be excluded from the optimisation. Defaults to all state variables except
for the first one.
}
\item{solution_type}{
If set to "eigen", the solution of the system of differential equations is
based on the spectral decomposition of the coefficient matrix in cases that
this is possible. If set to "deSolve", a numerical ode solver from package
\code{\link{deSolve}} is used. If set to "analytical", an analytical
solution of the model is used. This is only implemented for simple
degradation experiments with only one state variable, i.e. with no
metabolites. The default is "auto", which uses "analytical" if possible,
otherwise "eigen" if the model can be expressed using eigenvalues and
eigenvectors, and finally "deSolve" for the remaining models (time
dependence of degradation rates and metabolites). This argument is passed
on to the helper function \code{\link{mkinpredict}}.
}
\item{method.modFit}{
The optimisation method passed to \code{\link{modFit}}. The default "Marq"
is the Levenberg Marquardt algorithm \code{\link{nls.lm}} from the package
\code{minpack.lm}. Often other methods need more iterations to find the
same result. When using "Pseudo", "upper" and "lower" need to be
specified for the transformed parameters.
}
\item{control.modFit}{
Additional arguments passed to the optimisation method used by
\code{\link{modFit}}.
}
\item{plot}{
Should the observed values and the numerical solutions be plotted at each
stage of the optimisation?
}
\item{quiet}{
Suppress printing out the current model cost after each improvement?
}
\item{err }{either \code{NULL}, or the name of the column with the
\emph{error} estimates, used to weigh the residuals (see details of
\code{\link{modCost}}); if \code{NULL}, then the residuals are not weighed.
}
\item{weight}{
only if \code{err}=\code{NULL}: how to weight the residuals, one of "none",
"std", "mean", see details of \code{\link{modCost}}.
}
\item{scaleVar}{
Will be passed to \code{\link{modCost}}. Default is not to scale Variables
according to the number of observations.
}
\item{atol}{
Absolute error tolerance, passed to \code{\link{ode}}. Default is 1e-8,
lower than in \code{\link{lsoda}}.
}
\item{rtol}{
Absolute error tolerance, passed to \code{\link{ode}}. Default is 1e-10,
much lower than in \code{\link{lsoda}}.
}
\item{n.outtimes}{
The length of the dataseries that is produced by the model prediction
function \code{\link{mkinpredict}}. This impacts the accuracy of
the numerical solver if that is used (see \code{solution_type} argument.
The default value is 100.
}
\item{reweight.method}{
The method used for iteratively reweighting residuals, also known
as iteratively reweighted least squares (IRLS). Default is NULL,
the other method implemented is called "obs", meaning that each
observed variable is assumed to have its own variance, this is
estimated from the fit and used for weighting the residuals
in each iteration until convergence of this estimate up to
\code{reweight.tol} or up to the maximum number of iterations
specified by \code{reweight.max.iter}.
}
\item{reweight.tol}{
Tolerance for convergence criterion for the variance components
in IRLS fits.
}
\item{reweight.max.iter}{
Maximum iterations in IRLS fits.
}
\item{trace_parms}{
Should a trace of the parameter values be listed?
}
\item{\dots}{
Further arguments that will be passed to \code{\link{modFit}}.
}
}
\value{
A list with "mkinfit" and "modFit" in the class attribute.
A summary can be obtained by \code{\link{summary.mkinfit}}.
}
\note{
The implementation of iteratively reweighted least squares is inspired by the
work of the KinGUII team at Bayer Crop Science (Walter Schmitt and Zhenglei
Gao). A similar implemention can also be found in CAKE 2.0, which is the
other GUI derivative of mkin, sponsored by Syngenta.
}
\author{
Johannes Ranke <jranke@uni-bremen.de>
}
\examples{
# One parent compound, one metabolite, both single first order.
SFO_SFO <- mkinmod(
parent = list(type = "SFO", to = "m1", sink = TRUE),
m1 = list(type = "SFO"))
# Fit the model to the FOCUS example dataset D using defaults
fit <- mkinfit(SFO_SFO, FOCUS_2006_D)
summary(fit)
# Use stepwise fitting, using optimised parameters from parent only fit, FOMC
\dontrun{
FOMC <- mkinmod(parent = list(type = "FOMC"))
FOMC_SFO <- mkinmod(
parent = list(type = "FOMC", to = "m1", sink = TRUE),
m1 = list(type = "SFO"))
# Fit the model to the FOCUS example dataset D using defaults
fit.FOMC_SFO <- mkinfit(FOMC_SFO, FOCUS_2006_D)
# Use starting parameters from parent only FOMC fit
fit.FOMC = mkinfit(FOMC, FOCUS_2006_D, plot=TRUE)
fit.FOMC_SFO <- mkinfit(FOMC_SFO, FOCUS_2006_D,
parms.ini = fit.FOMC$bparms.ode, plot=TRUE)
}
# Use stepwise fitting, using optimised parameters from parent only fit, SFORB
SFORB <- mkinmod(parent = list(type = "SFORB"))
SFORB_SFO <- mkinmod(
parent = list(type = "SFORB", to = "m1", sink = TRUE),
m1 = list(type = "SFO"))
# Fit the model to the FOCUS example dataset D using defaults
fit.SFORB_SFO <- mkinfit(SFORB_SFO, FOCUS_2006_D)
# Use starting parameters from parent only SFORB fit (not really needed in this case)
fit.SFORB = mkinfit(SFORB, FOCUS_2006_D)
fit.SFORB_SFO <- mkinfit(SFORB_SFO, FOCUS_2006_D, parms.ini = fit.SFORB$bparms.ode, plot=TRUE)
# Weighted fits, including IRLS
SFO_SFO.ff <- mkinmod(parent = list(type = "SFO", to = "m1"),
m1 = list(type = "SFO"), use_of_ff = "max")
f.noweight <- mkinfit(SFO_SFO.ff, FOCUS_2006_D)
summary(f.noweight)
f.irls <- mkinfit(SFO_SFO.ff, FOCUS_2006_D, reweight.method = "obs")
summary(f.irls)
f.w.mean <- mkinfit(SFO_SFO.ff, FOCUS_2006_D, weight = "mean")
summary(f.w.mean)
f.w.mean.irls <- mkinfit(SFO_SFO.ff, FOCUS_2006_D, weight = "mean",
reweight.method = "obs")
summary(f.w.mean.irls)
# Manual weighting
dw <- FOCUS_2006_D
errors <- c(parent = 2, m1 = 1)
dw$err.man <- errors[FOCUS_2006_D$name]
f.w.man <- mkinfit(SFO_SFO.ff, dw, err = "err.man")
summary(f.w.man)
f.w.man.irls <- mkinfit(SFO_SFO.ff, dw, err = "err.man",
reweight.method = "obs")
summary(f.w.man.irls)
}
\keyword{ models }
\keyword{ optimize }
|
