APPLIED AND ENVIRONMENTAL MICROBIOLOGY, May 1983, p. 14531458
_{V}_{o}_{l}_{.} _{4}_{5}_{,} _{N}_{o}_{.} _{5}
00992240/83/05145306$02.00/0
_{C}_{o}_{p}_{y}_{r}_{i}_{g}_{h}_{t} C _{1}_{9}_{8}_{3}_{,} _{A}_{m}_{e}_{r}_{i}_{c}_{a}_{n} Society for Microbiology
Nonlinear Estimation of Monod Growth Kinetic Parameters
_{f}_{r}_{o}_{m} _{a} _{S}_{i}_{n}_{g}_{l}_{e} Substrate Depletion Curvet
JOSEPH A. ROBINSONt AND JAMES M. TIEDJE*
_{D}_{e}_{p}_{a}_{r}_{t}_{m}_{e}_{n}_{t} _{o}_{f} _{M}_{i}_{c}_{r}_{o}_{b}_{i}_{o}_{l}_{o}_{g}_{y} _{a}_{n}_{d} _{P}_{u}_{b}_{l}_{i}_{c} _{H}_{e}_{a}_{l}_{t}_{h}_{,} _{M}_{i}_{c}_{h}_{i}_{g}_{a}_{n} State University, East Lansing, Michigan 48824
Received 27 September 1982/Accepted 25 February 1983
Monod growth kinetic parameters were estimated by fitting sigmoidal substrate _{d}_{e}_{p}_{l}_{e}_{t}_{i}_{o}_{n} _{d}_{a}_{t}_{a} _{t}_{o} _{t}_{h}_{e} integrated Monod equation, using nonlinear leastsquares _{a}_{n}_{a}_{l}_{y}_{s}_{i}_{s}_{.} _{W}_{h}_{e}_{n} _{t}_{h}_{e} _{i}_{n}_{i}_{t}_{i}_{a}_{l} _{s}_{u}_{b}_{s}_{t}_{r}_{a}_{t}_{e} _{c}_{o}_{n}_{c}_{e}_{n}_{t}_{r}_{a}_{t}_{i}_{o}_{n} _{w}_{a}_{s} in the mixedorder region, _{n}_{o}_{n}_{l}_{i}_{n}_{e}_{a}_{r} _{e}_{s}_{t}_{i}_{m}_{a}_{t}_{i}_{o}_{n} _{o}_{f} _{s}_{i}_{m}_{u}_{l}_{a}_{t}_{e}_{d} _{d}_{a}_{t}_{a} _{s}_{e}_{t}_{s} containing known provided accurate estimates of the Px. _{K}_{s}_{,} and Y values used to create these data. Nonlinear regression analysis of sigmoidal substrate depletion data was also evaluated for H2limited batch growth of Desulfovibrio sp. strain Gil. The _{i}_{n}_{t}_{e}_{g}_{r}_{a}_{t}_{e}_{d} _{M}_{o}_{n}_{o}_{d} _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n} _{c}_{a}_{n} be more convenient for the estimation of growth kinetic _{p}_{a}_{r}_{a}_{m}_{e}_{t}_{e}_{r}_{s}_{,} particularly for gaseous substrates, but it must be recognized
measurement errors
that the estimates of
rate history of the inoculum.
Rma,x
_{K}_{5}_{,} and Y obtained may be influenced by the growth
_{B}_{o}_{t}_{h} derivative and integrated forms of equa tions derived for enzymecatalyzed reactions
_{h}_{a}_{v}_{e} _{b}_{e}_{e}_{n} used _{t}_{o} estimate kinetic _{p}_{a}_{r}_{a}_{m}_{e}_{t}_{e}_{r}_{s} for microbially mediated processes. In particu _{l}_{a}_{r}_{,} _{V}_{m}_{a} _{a}_{n}_{d} Km estimates have been calculated
_{b}_{y} _{f}_{i}_{t}_{t}_{i}_{n}_{g} data to either _{i}_{n}_{t}_{e}_{g}_{r}_{a}_{t}_{e}_{d} _{(}_{2}_{,} _{5}_{,} _{6}_{,} _{8}_{,} _{1}_{4}_{,} _{1}_{5}_{)} _{o}_{r} derivative _{(}_{3}_{,} 14, 15) forms of the Michae _{l}_{i}_{s}_{}_{M}_{e}_{n}_{t}_{e}_{n} _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n}_{.} _{H}_{o}_{w}_{e}_{v}_{e}_{r}_{,} MichaelisMen _{t}_{e}_{n} kinetics _{o}_{n}_{l}_{y} describe bacterial substrate
_{c}_{o}_{n}_{s}_{u}_{m}_{p}_{t}_{i}_{o}_{n} either _{(}_{i}_{)} when this _{p}_{r}_{o}_{c}_{e}_{s}_{s} is un
linked to growth, such as under resting condi _{t}_{i}_{o}_{n}_{s}_{,} _{o}_{r} _{(}_{i}_{i}_{)} when the amount of _{g}_{r}_{o}_{w}_{t}_{h} occur _{r}_{i}_{n}_{g} _{i}_{s} _{l}_{e}_{s}_{s} than that which _{g}_{i}_{v}_{e}_{s} _{s}_{i}_{g}_{m}_{o}_{i}_{d}_{a}_{l}
substrate _{d}_{e}_{p}_{l}_{e}_{t}_{i}_{o}_{n}_{.} _{V}_{m}_{.}_{,} is no longer a parame
if its value _{c}_{h}_{a}_{n}_{g}_{e}_{s} _{d}_{u}_{r}_{i}_{n}_{g}
ter _{(}_{i}_{.}_{e}_{.}_{,} a constant) substrate consumption.
When substrate _{c}_{o}_{n}_{s}_{u}_{m}_{p}_{t}_{i}_{o}_{n} is linked to
_{g}_{r}_{o}_{w}_{t}_{h}_{,} the number of catalytic
_{u}_{n}_{i}_{t}_{s}_{,} _{o}_{r} activi
_{t}_{y}_{,} increases with time.
_{A}_{s}_{s}_{u}_{m}_{i}_{n}_{g} thatthe initial
substrate concentration is greater than that
_{w}_{h}_{i}_{c}_{h} _{g}_{i}_{v}_{e}_{s} onehalf of the maximum growth
_{i}_{n}_{c}_{r}_{e}_{a}_{s}_{e} in _{a}_{c}_{t}_{i}_{v}_{i}_{t}_{y} concomitant with
rate, an
substrate consumption yields
an Sshaped sub
strate depletion curve, or sigmoidal kinetics
_{(}_{1}_{2}_{)}_{.} As mentioned _{a}_{b}_{o}_{v}_{e}_{,}
_{i}_{n}_{c}_{o}_{n}_{s}_{i}_{s}_{t}_{e}_{n}_{t} _{w}_{i}_{t}_{h} _{t}_{h}_{e} MichaelisMenten _{m}_{o}_{d}_{e}_{l}_{,}
sigmoidal kinetics is
but is predicted by
In this paper,
_{M}_{o}_{n}_{o}_{d} kinetics. we demonstrate the _{f}_{i}_{t}_{t}_{i}_{n}_{g} of
substrate depletion data (i.e., progress curve
_{d}_{a}_{t}_{a}_{)} to the integrated Monod equation by using
nonlinear regression,
which is advantageous
_{t} Journal article
Experiment Station.
no. 10579 of the Michigan Agricultural
_{t} _{P}_{r}_{e}_{s}_{e}_{n}_{t} address: _{D}_{e}_{p}_{a}_{r}_{t}_{m}_{e}_{n}_{t} of Civil Engineering, Mon
tana State University, Bozeman, _{M}_{T} 59717.
since estimates of growth kinetic parameters _{m}_{a}_{y} be _{c}_{a}_{l}_{c}_{u}_{l}_{a}_{t}_{e}_{d} from a single progress curve.
_{K}_{,} and
least
_{s}_{q}_{u}_{a}_{r}_{e}_{s} _{a}_{n}_{a}_{l}_{y}_{s}_{i}_{s}_{,} can be calculated _{b}_{y} using
_{l}_{i}_{n}_{e}_{a}_{r}_{i}_{z}_{e}_{d}_{,} discretized forms of the Monod equa _{t}_{i}_{o}_{n}_{s} _{d}_{e}_{s}_{c}_{r}_{i}_{b}_{i}_{n}_{g} the rates of _{c}_{h}_{a}_{n}_{g}_{e} of biomass
substrate concentrations
batch _{g}_{r}_{o}_{w}_{t}_{h}_{.} Additionally, we discuss
_{t}_{h}_{e} _{u}_{s}_{e} _{o}_{f} sensitivity coefficients for the optimal
estimation of growth kinetic parameters and the advantages of using the integrated Monod equa tion for estimation of Lmax, K, and Y. Finally, _{w}_{e} _{g}_{i}_{v}_{e} _{a}_{t}_{t}_{e}_{n}_{t}_{i}_{o}_{n} to the limitations of fitting
during
and growthlimiting
_{Y}_{,} _{w}_{h}_{i}_{c}_{h} _{a}_{r}_{e} required for nonlinear
We show how initial estimates of _{U}_{m}_{a}_{,}
transientgrowth data to the Monod model,
_{w}_{h}_{i}_{c}_{h} _{d}_{e}_{s}_{c}_{r}_{i}_{b}_{e}_{s} balanced bacterial growth only
(13).
_{T}_{H}_{E}_{O}_{R}_{Y} AND METHODS The rate of _{c}_{h}_{a}_{n}_{g}_{e} of substrate consumption by a
bacterium _{g}_{r}_{o}_{w}_{i}_{n}_{g} in batch _{m}_{a}_{y} be described _{b}_{y}
dS/dt = [ m,,.,,S/(K5 + S)]X/Y
(1)
where _{,}_{u}_{,} is the maximum specific growth rate, K5 is
growth, and Y is the
yield
centration and _{m}_{a}_{y} be eliminated from _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n} 1 by
using
X = Y(SO  S) + XO
(2)
_{w}_{h}_{i}_{c}_{h} _{r}_{e}_{l}_{a}_{t}_{e}_{s} _{X} _{a}_{t} _{t}_{i}_{m}_{e} _{t} _{t}_{o} _{S}_{.} After elimination of _{X}_{,}
coefficient. The variable X is the biomass con
the halfsaturation constant for
equation 1 becomes
dS/dt = _{}_{[}_{p}_{L}_{,}_{,}_{S}_{/}_{(}_{K}_{,} + S)][Y(So  S) + Xo)]/Y (3)
which _{m}_{a}_{y} be integrated to give
Cjln{[Y(SO  S) + XO]/XO}  C21n(S/So} = Lmaxt(4)
1453
1454 ROBINSON _{A}_{N}_{D} _{T}_{I}_{E}_{D}_{J}_{E}
where C1 = (KSY + SOY + XO)/(YSO +
K5Y/(YSo + _{X}_{O}_{)}_{.} Equation 4 gives the familiar _{S}_{}
and _{C}_{2} _{=}
XO)
shaped curve for substrate
growth. A relation similar to
increase of _{b}_{i}_{o}_{m}_{a}_{s}_{s}
derived
depletion
equation
batch
during batch
describing the
4
has _{b}_{e}_{e}_{n}
obtained _{a}_{f}_{t}_{e}_{r} using the _{m}_{a}_{s}_{s}
solved for _{S}_{,} _{a}_{n}_{d}
growth
approximated.
may
during
by integrating the expression
S from
equation 1 _{b}_{y}
eliminating
balance relation _{(}_{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n}
Equation 4 _{c}_{a}_{n}_{n}_{o}_{t} be hence its solution must _{b}_{e}
2) (11, 12). explicitly
numerically
_{m}_{a}_{y}
Solution curves _{(}_{S} _{v}_{e}_{r}_{s}_{u}_{s} _{t}_{)}
solving equation 3 _{b}_{y} numerical
be estimated _{b}_{y}
_{o}_{r}_{,} alter
be
integration,
and X
nately, solution curves for both
S
generated by solving equation 1 simultaneously with
dX/dt = [~z,u.S/(K, + S)]X
(5)
again by numerical
integration _{(}_{4}_{)}_{.}
_{a}_{n}_{a}_{l}_{y}_{s}_{i}_{s}
variable to
be
Nonlinear regression sitivity of the dependent
ment (although these
Y
may
_{r}_{e}_{q}_{u}_{i}_{r}_{e}_{s} _{t}_{h}_{a}_{t} _{t}_{h}_{e} _{s}_{e}_{n}_{}
changes
in each _{o}_{f}
derivatives of _{S}
this require
approximat
derived from _{t}_{h}_{e}
the parameters be _{c}_{a}_{l}_{c}_{u}_{l}_{a}_{b}_{l}_{e}_{.} _{T}_{h}_{e} _{f}_{i}_{r}_{s}_{t}
with respect to p. _{,}_{s} _{K}_{5}_{,} _{a}_{n}_{d} _{Y}
satisfy
numerically
ed), and these
integrated Monod
ation (16). The
expressions can be
equation
by
sensitivity equations
using _{i}_{m}_{p}_{l}_{i}_{c}_{i}_{t} differenti
for tmax, _{K}_{S}_{,} and
are _{g}_{i}_{v}_{e}_{n} below.
dS/dR,max = t/C4
(6)
dS/dK, = {(Y/C3}Rln(X/XO}  ln(S/So}]}/C4 (7)
dS/dY =
{C1(S0  S)/X +
C1)SO]
ln(X/XO}/C3[K,
C2SO)}/C4
 ln(S/SO)/C3(K,

+
(1

(8)
In
the above three
equations the terms _{C}_{3} and _{C}_{4} _{e}_{q}_{u}_{a}_{l}
C4/S,
respectively. _{N}_{o}_{t}_{e} _{t}_{h}_{a}_{t}
By
is
of _{K}_{S} and _{Y}_{.}
equation
YSO + XO and C3Y/X +
equations _{6}_{,} _{7}_{,} and 8 are all functions
definition, _{t}_{h}_{e}_{n}_{,} the
nonlinear,
integrated Monod
since its sensitivity
equations are not inde
for nonlinear
pendent of _{t}_{h}_{e} In addition _{t}_{o}
parameters _{(}_{1}_{)}_{.}
being required
regres
sion, the sensitivity _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n}_{s} _{p}_{r}_{e}_{d}_{i}_{c}_{t} _{(}_{i}_{)} _{w}_{h}_{e}_{t}_{h}_{e}_{r} _{t}_{h}_{e}
parameters _{m}_{a}_{y} _{b}_{e}
uniquely estimated, _{(}_{i}_{i}_{)} _{t}_{h}_{e} _{r}_{e}_{l}_{a}_{t}_{i}_{v}_{e}
parameters, _{a}_{n}_{d} _{(}_{i}_{i}_{i}_{)} _{t}_{h}_{e}
independent _{v}_{a}_{r}_{i}_{a}_{b}_{l}_{e} _{o}_{v}_{e}_{r} _{w}_{h}_{i}_{c}_{h} _{t}_{h}_{e}
precision of the estimated
range of _{t}_{h}_{e}
model is most sensitive to
The last item is useful in
ments _{f}_{o}_{r}
changes
in the
parameters.
designing _{o}_{p}_{t}_{i}_{m}_{a}_{l} _{e}_{x}_{p}_{e}_{r}_{i}_{}
parameter estimation _{(}_{1}_{)}_{.} _{I}_{f} the sensitivity
multiples of
then it is impossible to _{o}_{b}_{t}_{a}_{i}_{n} _{u}_{n}_{i}_{q}_{u}_{e}
equations are proportional _{(}_{i}_{.}_{e}_{.}_{,} _{i}_{f} _{t}_{h}_{e}_{y} _{a}_{r}_{e}
one another),
data
similar, but _{t}_{h}_{e}_{y} _{a}_{r}_{e} _{h}_{i}_{g}_{h}_{l}_{y}
of parameter estimates
set, and this is _{t}_{r}_{u}_{e} _{f}_{o}_{r}
(Fig.
estimates _{o}_{f} _{t}_{h}_{e} parameters _{f}_{r}_{o}_{m} _{t}_{h}_{e}
least
squares analysis _{(}_{1}_{)}_{.} Unique parameter estimates _{m}_{a}_{y}
by
be obtained _{w}_{h}_{e}_{n} _{t}_{h}_{e} sensitivity _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n}_{s} _{a}_{r}_{e} _{v}_{e}_{r}_{y}
correlated. _{T}_{h}_{i}_{s} _{s}_{i}_{t}_{u}_{a}_{t}_{i}_{o}_{n} _{i}_{s}
undesirable
since it implies that several combinations
may describe the _{s}_{a}_{m}_{e} _{d}_{a}_{t}_{a}
equation 4, depending _{o}_{n} _{S}_{O}_{.}
theone forKS.
The sensitivity _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n}_{s} _{f}_{o}_{r} _{I}_{m}_{a}_{x}_{,} _{K}_{S}_{,} _{a}_{n}_{d} _{Y} _{y}_{i}_{e}_{l}_{d}
sensitivity _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n}_{s}
similar _{c}_{u}_{r}_{v}_{e}_{s}
for _{m}_{.}_{L}_{m}_{a}_{x} and Y numericallydominating
_{1}_{)}_{,} _{w}_{i}_{t}_{h} _{t}_{h}_{e}
In the firstorder _{r}_{e}_{g}_{i}_{o}_{n} _{i}_{t} _{i}_{s} _{n}_{o}_{t} _{p}_{o}_{s}_{s}_{i}_{b}_{l}_{e} _{t}_{o} _{o}_{b}_{t}_{a}_{i}_{n}
unique estimates _{o}_{f} _{m}
sensitivity coefficients _{f}_{o}_{r} _{L}_{m}_{a}_{x}_{a}_{n}_{d} _{Y} areproportional
(Fig. 1A). This is intuitively _{c}_{l}_{e}_{a}_{r} _{s}_{i}_{n}_{c}_{e} _{a}_{n} _{i}_{n}_{c}_{r}_{e}_{a}_{s}_{e} _{i}_{n}
_{K}_{S}_{,} _{a}_{n}_{d} _{Y} _{b}_{e}_{c}_{a}_{u}_{s}_{e} _{t}_{h}_{e}
IL,,I is equivalent to a decrease in Y in _{t}_{h}_{e} _{f}_{i}_{r}_{s}_{t}_{}_{o}_{r}_{d}_{e}_{r}
region. The sensitivity _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n}_{s} _{a}_{l}_{s}_{o} _{p}_{r}_{e}_{d}_{i}_{c}_{t} _{t}_{h}_{a}_{t}
relatively _{p}_{o}_{o}_{r} _{e}_{s}_{t}_{i}_{m}_{a}_{t}_{e}_{s} _{o}_{f} _{.}_{,}_{m}_{a}_{x}_{,} _{K}_{S}_{,} _{a}_{n}_{d} _{Y} _{w}_{i}_{l}_{l} _{b}_{e}
obtained when So is saturating.
This results _{f}_{r}_{o}_{m} _{t}_{h}_{e}
APPL. ENVIRON. MICROBIOL.
parameters
over most _{o}_{f}
_{S}_{o} for the optimal
having _{n}_{e}_{a}_{r}_{l}_{y} _{p}_{r}_{o}_{p}_{o}_{r}_{t}_{i}_{o}_{n}_{a}_{l} sensitivities
the
progress curve
(Fig.
1C). The value of
estimation _{o}_{f}
_{r}_{e}_{g}_{i}_{o}_{n}
_{(}_{F}_{i}_{g}_{.} _{1}_{B}_{)}_{.}
_{m}_{a}_{y}
techniques
but
we
_{L}_{m}_{a}_{x}_{,} _{K}_{S}_{,} _{a}_{n}_{d} _{Y} _{i}_{s} _{i}_{n}
be used to fit data _{t}_{o}
chose the Gaussian _{m}_{e}_{t}_{h}_{o}_{d}
simplicity _{(}_{1}_{)}_{.} The _{G}_{a}_{u}_{s}_{s}_{i}_{a}_{n}
the mixedorder
A number of nonlinear _{m}_{o}_{d}_{e}_{l}_{s}_{,}
because of its relative
method uses the equation
S  ST = AlXraxdS/dp.max + _{A}_{K}_{5}_{d}_{S}_{/}_{d}_{K}_{,} + _{A}_{Y}_{d}_{S}_{/}_{d}_{Y}
(9)
where _{S}_{T} _{i}_{S} _{t}_{h}_{e} centration _{a}_{t} _{t}_{i}_{m}_{e}
t
and AILmax, AK,,and
parameters.
The
the
the
evaluating
calculating
positive
theoretically
predicted substrate con
_{p}_{a}_{r}_{a}_{m}_{e}_{t}_{e}_{r} estimates,
proceeds
by
givencurrent
AY are
application
sensitivity
residual
correction terms for these
of
6, 7, and 8 and
equation 9
equations
errors
_{(}_{S}_{T}  _{S}_{)} at the _{m}_{e}_{a}_{}
parame
_{p}_{a}_{r}_{a}_{m}_{e}_{t}_{e}_{r} estimates.
some
ILmax,
KS,
sured
terestimates. _{T}_{h}_{e} _{c}_{o}_{r}_{r}_{e}_{c}_{t}_{i}_{o}_{n} _{t}_{e}_{r}_{m}_{s} _{a}_{r}_{e} _{t}_{h}_{e}_{n} _{c}_{a}_{l}_{c}_{u}_{l}_{a}_{t}_{e}_{d}
and added (they can be
via
then serve _{a}_{s} _{t}_{h}_{e} and this
process
These
initialestimates _{f}_{o}_{r}_{t}_{h}_{e} _{n}_{e}_{x}_{t}
continues untilthe correction terms _{a}_{r}_{e} _{l}_{e}_{s}_{s} _{t}_{h}_{a}_{n}
small value
_{0}_{.}_{0}_{1}_{)}_{.} Once the solution _{h}_{a}_{s} _{c}_{o}_{n}_{}
and _{Y}
_{m}_{a}_{y} be calculated _{f}_{r}_{o}_{m} _{t}_{h}_{e} _{v}_{a}_{r}_{i}_{a}_{n}_{c}_{e}_{s} _{o}_{f} _{t}_{h}_{e}_{i}_{r}
substrate concentrations _{f}_{o}_{r} _{t}_{h}_{e} _{i}_{n}_{i}_{t}_{i}_{a}_{l}
multiple linear
or
negative)
regression
to the initial
parameter
corrected
(e.g.,
estimates
iteration,
_{v}_{e}_{r}_{g}_{e}_{d}_{,} estimates
respective
of the standard _{e}_{r}_{r}_{o}_{r}_{s} _{o}_{f}
(8).
correction terms
Nonlinear nonlinear in _{i}_{t}_{s}
regression _{a}_{n}_{a}_{l}_{y}_{s}_{i}_{s} for _{a}_{n}_{y} _{m}_{o}_{d}_{e}_{l} _{t}_{h}_{a}_{t} _{i}_{s}
parameters requires initial _{e}_{s}_{t}_{i}_{m}_{a}_{t}_{e}_{s} _{o}_{r}
"'guesses"
done
the
of
the parameters _{(}_{1}_{,} _{7}_{)}_{.} This is
usually
by fitting
model,
transformed _{d}_{a}_{t}_{a} _{t}_{o} _{a} _{l}_{i}_{n}_{e}_{a}_{r} _{v}_{e}_{r}_{s}_{i}_{o}_{n}
but the integrated Monod equation is
(7) and cannot be transformed
of
intrinsically _{n}_{o}_{n}_{l}_{i}_{n}_{e}_{a}_{r}
into a linear expression for the purpose of estimating
P'max,
KS, and Y.
However,
be
provisional
estimates _{o}_{f}
these parameters _{m}_{a}_{y}
obtained by
using
AS/At = [(LmaxS/(Ks + _{S}_{)}_{]}_{X}_{/}_{Y} AX/At = [j±maxS/(Ks + _{S}_{)}_{]}_{X}
(10)
(11)
Equations 10 and 11 are derived by replacing infinitesi
mal
time(dt)withfinitetime _{(}_{A}_{t}_{)}_{a}_{n}_{d}
ly
correct if At is
are approximate
relatively small _{(}_{1}_{6}_{)}_{.} These _{f}_{i}_{n}_{i}_{t}_{e}_{}
difference _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n}_{s} are nonlinear, _{b}_{u}_{t} _{m}_{a}_{y} _{b}_{e}
converted _{i}_{n}_{t}_{o} the following linearized _{f}_{o}_{r}_{m}_{s}_{:}
AtX/AS = (KsY4lumax)(1/S) + _{Y}_{/}_{I}_{m}_{a}_{x} _{(}_{1}_{2}_{)}
and
_{(}_{1}_{3}_{)}
which yield straight _{l}_{i}_{n}_{e}_{s} _{w}_{h}_{e}_{n} _{}_{A}_{t}_{X}_{/}_{A}_{S} _{a}_{n}_{d} _{A}_{t}_{X}_{/}_{A}_{X}
are plotted against 1/S.
AtX/AX = (Ks4Imax)(1/S) + _{l}_{4}_{l}_{m}_{a}_{x}
RESULTS AND DISCUSSION
To evaluate nonlinear
regression _{a}_{n}_{a}_{l}_{y}_{s}_{i}_{s} _{f}_{o}_{r}
data containing
equation _{4}_{,} we fitted _{s}_{i}_{m}_{u}_{l}_{a}_{t}_{e}_{d}
known errors _{t}_{o} _{t}_{h}_{i}_{s}
ian method. Theoretically, _{e}_{s}_{t}_{i}_{m}_{a}_{t}_{e}_{s} _{o}_{f} _{I}_{m}_{a}_{x}_{,}
equation, _{u}_{s}_{i}_{n}_{g} _{t}_{h}_{e} _{G}_{a}_{u}_{s}_{s}_{}
_{K}_{,} and Y should be close _{t}_{o} _{t}_{h}_{o}_{s}_{e} parameter
values used to generate the simulated _{d}_{a}_{t}_{a}_{.} _{B}_{y}
analyzing simulated _{d}_{a}_{t}_{a} _{i}_{t} _{i}_{s} _{p}_{o}_{s}_{s}_{i}_{b}_{l}_{e} _{t}_{o} _{c}_{h}_{e}_{c}_{k}
whether the sensitivity coefficients _{h}_{a}_{v}_{e} _{b}_{e}_{e}_{n}
VOL. _{4}_{5}_{,} 1983
A
B
C
0
la 0
'a
0
a
0
X
'0
10
0
0
_{o} _{0} '0
NONLINEAR REGRESSION _{A}_{N}_{A}_{L}_{Y}_{S}_{I}_{S} _{O}_{F} _{M}_{O}_{N}_{O}_{D} _{D}_{A}_{T}_{A}
_{1}_{4}_{5}_{5}
between these values, and provisional
linear
ized, discretized _{v}_{e}_{r}_{s}_{i}_{o}_{n} _{o}_{f} _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n} _{3}_{.}_{)} _{T}_{o} _{f}_{i}_{t}
data to equations 12 and 13, estimates _{o}_{f} _{}_{d}_{S}_{/}_{d}_{t}
and dX/dt are required. _{T}_{h}_{e}_{s}_{e} _{w}_{e}_{r}_{e} _{o}_{b}_{t}_{a}_{i}_{n}_{e}_{d}
fitting the St _{a}_{n}_{d} _{X}_{}_{t} _{d}_{a}_{t}_{a} _{p}_{a}_{i}_{r}_{s} _{t}_{o} _{c}_{u}_{b}_{i}_{c} _{s}_{p}_{l}_{i}_{n}_{e}_{s}
by
(4) and then evaluating the first derivatives _{o}_{f}
examples _{o}_{f}
fitting transformed data _{h}_{a}_{v}_{i}_{n}_{g} _{e}_{i}_{t}_{h}_{e}_{r} _{r}_{e}_{l}_{a}_{t}_{i}_{v}_{e} _{o}_{r}
simple measurement _{e}_{r}_{r}_{o}_{r}_{s} _{a}_{r}_{e}
the fitted cubic polynomials. Two
Ks estimates can be calculated by _{u}_{s}_{i}_{n}_{g} _{a}
pm. and
depicted in _{F}_{i}_{g}_{.}
pmae were
of the
2. For both cases (A _{a}_{n}_{d} _{B}_{)}_{,} _{K}_{s} _{a}_{n}_{d}
calculated from the y intercepts _{a}_{n}_{d} _{s}_{l}_{o}_{p}_{e}_{s}
straight lines _{f}_{i}_{t}_{t}_{e}_{d} _{t}_{o} _{t}_{h}_{e} _{t}_{r}_{a}_{n}_{s}_{f}_{o}_{r}_{m}_{e}_{d} _{b}_{i}_{o}_{m}_{a}_{s}_{s}
data. The two Y estimates were calculated by dividing the y intercepts _{o}_{f} _{t}_{h}_{e} _{b}_{e}_{s}_{t}_{}_{f}_{i}_{t} _{l}_{i}_{n}_{e}_{s} _{f}_{o}_{r} transformed _{s}_{u}_{b}_{s}_{t}_{r}_{a}_{t}_{e} _{d}_{a}_{t}_{a} _{b}_{y} _{t}_{h}_{e} _{y} _{i}_{n}_{t}_{e}_{r}_{c}_{e}_{p}_{t}_{s} _{o}_{f} the linear equations _{f}_{i}_{t}_{t}_{e}_{d} _{t}_{o} _{t}_{h}_{e} _{t}_{r}_{a}_{n}_{s}_{f}_{o}_{r}_{m}_{e}_{d}
biomass _{d}_{a}_{t}_{a}_{.}
The initial estimates of _{i}_{u}_{m}_{.}_{,} _{K}_{5}_{,} _{a}_{n}_{d} _{Y} _{w}_{e}_{r}_{e}
entered into a computer _{p}_{r}_{o}_{g}_{r}_{a}_{m} _{(}_{M}_{O}_{N}_{}
ODCRV) that fits _{S}
Gaussian _{m}_{e}_{t}_{h}_{o}_{d}_{.} _{I}_{n} addition to estimating the
data to equation _{4} _{b}_{y} _{t}_{h}_{e}
MONODCRV _{a}_{p}_{}
growth kinetic parameters,
proximates _{(}_{i}_{)} _{t}_{h}_{e} _{s}_{t}_{a}_{n}_{d}_{a}_{r}_{d} _{e}_{r}_{r}_{o}_{r}_{s} _{o}_{f} _{F}_{L}_{m}_{.}_{,} _{K}_{5}_{,}
and Y, assuming no correlation _{a}_{m}_{o}_{n}_{g} _{m}_{e}_{a}_{s}_{u}_{r}_{e}_{} ment errors, and (ii) the absolute residual _{s}_{u}_{m}_{} ofsquares at each iteration. The _{p}_{a}_{t}_{t}_{e}_{r}_{n} of _{p}_{a}_{}
Time
FIG. 1. _{S}_{e}_{n}_{s}_{i}_{t}_{i}_{v}_{i}_{t}_{y} coefficients _{f}_{o}_{r} _{P}_{r}_{a}_{x}
(dS/diL,,), _{K}_{s} _{(}_{d}_{S}_{/}_{d}_{K}_{5}_{)}_{,}
and Y (dS/dY). For all three
panels, _{A}_{,}_{,}_{,} = _{0}_{.}_{1}_{,} _{K}_{s} = _{5}_{,} Y
= _{0}_{.}_{2}_{,} _{a}_{n}_{d} _{X}_{O} = _{1}_{.} _{S}_{O} _{=}
0.1, 20, and 250 for panels A, B, and C, respectively.
correctly _{d}_{e}_{r}_{i}_{v}_{e}_{d}_{,} since incorrect
pressions can _{(}_{i}_{)} _{p}_{r}_{e}_{v}_{e}_{n}_{t} _{f}_{i}_{t}_{t}_{i}_{n}_{g} _{d}_{a}_{t}_{a} _{t}_{o} _{a} _{g}_{i}_{v}_{e}_{n} model altogether or _{(}_{i}_{i}_{)} lead _{t}_{o} _{d}_{r}_{a}_{m}_{a}_{t}_{i}_{c} _{e}_{r}_{r}_{o}_{r}_{s} _{i}_{n} the estimated parameters. _{T}_{w}_{e}_{l}_{v}_{e} _{s}_{i}_{m}_{u}_{l}_{a}_{t}_{e}_{d}
sensitivity ex
data sets _{w}_{e}_{r}_{e} _{c}_{r}_{e}_{a}_{t}_{e}_{d} _{b}_{y}
numerically integrat
ingequation 2forthefollowingparameter _{v}_{a}_{l}_{u}_{e}_{s}
and initial conditions: _{m}
0.2, _{S}_{o} = 20, and _{X}_{o} = 1. Measurement _{e}_{r}_{r}_{o}_{r}_{s} _{o}_{f} the relative type (constant coefficient _{o}_{f} _{v}_{a}_{r}_{i}_{a}_{}
_{t}_{i}_{o}_{n}_{)} were added to _{s}_{i}_{x} _{o}_{f} _{t}_{h}_{e} _{e}_{r}_{r}_{o}_{r}_{}_{f}_{r}_{e}_{e} _{d}_{a}_{t}_{a}
sets, using a pseudorandom number
according to the _{p}_{r}_{o}_{c}_{e}_{d}_{u}_{r}_{e} described _{b}_{y} _{H}_{a}_{r}_{} _{b}_{a}_{u}_{g}_{h} and BonhamCarter _{(}_{1}_{0}_{)}_{.} _{S}_{i}_{m}_{i}_{l}_{a}_{r}_{l}_{y}_{,} _{s}_{i}_{m}_{}
ple errors _{(}_{c}_{o}_{n}_{s}_{t}_{a}_{n}_{t} standard _{d}_{e}_{v}_{i}_{a}_{t}_{i}_{o}_{n}_{)} _{w}_{e}_{r}_{e} introduced into the _{r}_{e}_{m}_{a}_{i}_{n}_{i}_{n}_{g} _{s}_{i}_{x} _{d}_{a}_{t}_{a} _{s}_{e}_{t}_{s}_{.}
generator
= _{0}_{.}_{1}_{,} _{K}_{s} = _{5}_{,} _{Y} =
Initial
,
_{K}_{5}_{,} and Y _{e}_{s}_{t}_{i}_{m}_{a}_{t}_{e}_{s} _{w}_{e}_{r}_{e} _{o}_{b}_{}
tained from linear regression _{a}_{n}_{a}_{l}_{y}_{s}_{i}_{s} of the
simulated St and Xt data _{p}_{a}_{i}_{r}_{s} _{t}_{r}_{a}_{n}_{s}_{f}_{o}_{r}_{m}_{e}_{d}
according to equations 12 and 13. (Actually,
_{o}_{n}_{l}_{y} initial and final estimates of the _{b}_{i}_{o}_{m}_{a}_{s}_{s}
concentration are _{r}_{e}_{q}_{u}_{i}_{r}_{e}_{d}_{;} a _{p}_{r}_{o}_{v}_{i}_{s}_{i}_{o}_{n}_{a}_{l} esti
mate of Y can be obtained from the _{d}_{i}_{f}_{f}_{e}_{r}_{e}_{n}_{c}_{e}
A 
An 

so6 

0 

la 40 

20 

0 

o 
2 

I 

^{B} 
12 

9I 

Go 

10 

x 

0 

3 

0 

0 
0.2 
400 

0 

300 

x 

200x 

SD=0.01 
100 

_{4} 
_{6} 
a 

1/8 

ist~~~~~~~~~~~~~~~~~~~~ 

0 Uvv 
x
a
cv 0.005
_{0}_{.}_{4}
118
0.6
40
30X
x
20 0
10
n
0.8
FIG. 2. Linearized, discretized Monod data. _{S}_{i}_{m}_{u}_{}
lated data containing errors _{h}_{a}_{v}_{i}_{n}_{g} a constant coeffi
cient of variation _{(}_{C}_{V}_{)} _{o}_{f} _{0}_{.}_{5}_{%} _{(}_{A}_{)} _{a}_{n}_{d} _{a} _{c}_{o}_{n}_{s}_{t}_{a}_{n}_{t} standard deviation (SD) of 0.01 _{(}_{B}_{)} were transformed
according to _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n}_{s} 12 and 13. The values of _{l}_{m}_{a}_{x}_{,} _{K}_{5}_{,} and Y are the same _{a}_{s} those _{u}_{s}_{e}_{d} _{i}_{n} _{F}_{i}_{g}_{.} _{1}_{.}
1456 ROBINSON AND TIEDJE
rameter updating for data fitted to equation 4 is illustrated in Table 1. In both cases, five itera
_{t}_{i}_{o}_{n}_{s} _{w}_{e}_{r}_{e} _{r}_{e}_{q}_{u}_{i}_{r}_{e}_{d} before the correction _{t}_{e}_{r}_{m}_{s}
_{w}_{e}_{r}_{e} _{l}_{e}_{s}_{s} _{t}_{h}_{a}_{n} _{0}_{.}_{0}_{1} _{a}_{n}_{d} _{t}_{h}_{e} residual sumof
_{s}_{q}_{u}_{a}_{r}_{e}_{s} _{w}_{a}_{s} minimized. The _{u}_{p}_{d}_{a}_{t}_{i}_{n}_{g} of initial
parameter estimates was usually complete after _{f}_{o}_{u}_{r} or five iterations when a _{c}_{o}_{n}_{v}_{e}_{r}_{g}_{e}_{n}_{c}_{e} crite rion of 0.01 was used. Nonlinear regression analysis of the simulated Monod data generally provided better estimates of _{F}_{L}_{m}_{a}_{x}_{,} _{K}_{s}_{,} and Y than did leastsquares analy sis of the linearized data _{(}_{T}_{a}_{b}_{l}_{e} _{2}_{,} _{F}_{i}_{g}_{.} _{3}_{)}_{.} This was always true for _{K}_{s}_{,} but in a few cases the final _{R}_{m}_{a}_{x} (simulations 4 and 9) and Y (simula tion 5) estimates were slightly more in error than the initial estimates of these two parameters. Nonlinear regression analysis always increased the overall goodnessoffit by _{r}_{e}_{d}_{u}_{c}_{i}_{n}_{g} the resid ual sumofsquares, in some cases by as much as 104fold (simulation 9). In addition to the results presented above, we have found that fittingdata to equation 4 by using nonlinear regression analysis is far superior to linear analysis of transformed data for several combinations of _{p}_{a}_{r}_{a}_{m}_{e}_{t}_{e}_{r} values and error levels. _{H}_{o}_{w}_{e}_{v}_{e}_{r}_{,} when the relative error level is _{g}_{r}_{e}_{a}_{t}_{e}_{r} than _{o}_{r}
_{e}_{q}_{u}_{a}_{l} to _{5}_{%}_{,} then estimates of _{i}_{J}_{m}_{a}_{x}_{,} _{K}_{s}_{,} _{a}_{n}_{d} Y
may be in substantial error (e.g., 500%), al though the residual sumofsquares has been
minimized.
To demonstrate the fitting of biological data _{t}_{o} equation 4 by using nonlinear regression, _{w}_{e} obtained H2 depletion and biomass formation data for the sulfate reducer Desulfovibrio sp. strain Gl1 growing on H2 as the sole electron donor. The culture conditions and experimental details will be described elsewhere (Robinson
APPL. ENVIRON. MICROBIOL.
and Tiedje, manuscript in preparation). The growth of this organism was monitored by fol lowing the protein content. Initial estimates of the growth kinetic _{p}_{a}_{r}_{a}_{m}_{e}_{}
ters were calculated according to the above described procedure, and these were updated by MONODCRV. The goodnessoffit for the H2 _{c}_{o}_{n}_{c}_{e}_{n}_{t}_{r}_{a}_{t}_{i}_{o}_{n} data versus the theoretical _{c}_{u}_{r}_{v}_{e}
_{c}_{a}_{l}_{c}_{u}_{l}_{a}_{t}_{e}_{d} _{f}_{r}_{o}_{m} the best _{p}_{a}_{r}_{a}_{m}_{e}_{t}_{e}_{r} estimates _{i}_{s}
shown in Fig. 4; for these data, estimates of
lLmax,
Ks,
h1, 2.4
and Y,
respectively, were 5.6 x 102
and 0.99 _{g} of _{p}_{r}_{o}_{t}_{e}_{i}_{n}
jxM dissolved
H2,
per mol of H2. _{O}_{t}_{h}_{e}_{r}_{s} have determined bacterial _{g}_{r}_{o}_{w}_{t}_{h} _{p}_{a}_{} rameters from batch data, but generally they have used computationally inefficient _{p}_{r}_{o}_{c}_{e}_{} _{d}_{u}_{r}_{e}_{s} _{o}_{r} _{p}_{a}_{i}_{d} little attention to the limitations of
estimating these parameters under transient or
nonsteadystate conditions. Graham and Canale (9), in a recent investigation on the kinetics of a fourtrophic level predatorprey _{s}_{y}_{s}_{t}_{e}_{m}_{,} fitted
_{b}_{a}_{t}_{c}_{h} _{d}_{a}_{t}_{a} to Monod _{g}_{r}_{o}_{w}_{t}_{h} models without _{c}_{o}_{n}_{s}_{i}_{d}_{e}_{r}_{i}_{n}_{g} the influence that the _{g}_{r}_{o}_{w}_{t}_{h} rate
histories of their predator and _{p}_{r}_{e}_{y} cultures might have had on the optimal parameter esti mates. Before this, investigators typically fitted
biomass formation data to the _{i}_{n}_{t}_{e}_{g}_{r}_{a}_{t}_{e}_{d} version
of the Monod equation describing biomass _{p}_{r}_{o}_{} duction, using a systematic procedure (11) which involves changing _{P}_{t}_{m}_{a}_{x}_{,} _{K}_{s}_{,} and Y until the theoretical biomass formation _{c}_{u}_{r}_{v}_{e} _{a}_{g}_{r}_{e}_{e}_{s} with the experimental points. Systematic (or random) search techniques are inefficient and should be avoided if possible (1). The larger danger, however, is that since sensitivity coeffi cients are not required for systematic fitting procedures, the researcher may believe that it is
TABLE 1. Course of parameter updating during nonlinear regression analysis of simulated _{M}_{o}_{n}_{o}_{d} _{d}_{a}_{t}_{a}_{a}
_{E}_{s}_{t}_{i}_{m}_{a}_{t}_{e}_{b} 
Relative 
error' 
Simple error' 

IJLmax 
K5 
Y 
RSS' 
ILmax 
Ks 
Y 
RSS 

Initialf 
0.14 
10.5 
0.31 
80.9 
0.11 
5.99 
0.18 
46.70 
Iteration 1 
0.12 
7.26 
0.21 
8.9 
0.12 
7.40 
0.21 
0.83 
Iteration 2 
0.10 
4.58 
0.20 
1.1 
0.10 
4.49 
0.20 
0.16 
Iteration 3 
0.10 
5.02 
0.21 
0.09 
0.10 
5.00 
0.20 
0.004 
Iteration 4 
0.10 
4.95 
0.21 
0.08 
0.10 
5.04 
0.20 
0.003 
Final (iteration 5) 
0.10 
4.95 
0.21 
0.08 
0.10 
5.04 
0.20 
0.003 
True value 
0.10 
5.00 
0.20 
0.10 
5.00 
0.20 
a Untransformed data were directly fitted to equation 4 by using nonlinear regression analysis, given the initial _{p}_{a}_{r}_{a}_{m}_{e}_{t}_{e}_{r} estimates.
passes were required before the solution converged. For these simulated data, a
convergence
b
Generally, four or _{f}_{i}_{v}_{e}
criterion of 0.01 was used.
c The relative errors have a constant coefficient of variation (0.5%).
d The simple errors have a constant standard deviation (0.01).
e RSS, Residual sumofsquares.
f Initial estimates of fLmax, _{K}_{s}_{,} and Y were calculated by fitting transformed St and Xt data pairs to equations 12 and 13.
VOL. _{4}_{5}_{,} 1983
_{N}_{O}_{N}_{L}_{I}_{N}_{E}_{A}_{R} REGRESSION ANALYSIS OF MONOD DATA
1457
_{T}_{A}_{B}_{L}_{E} 2. Comparison
Simulation
_{o}_{f} _{e}_{r}_{r}_{o}_{r}_{s} _{i}_{n} _{p}_{a}_{r}_{a}_{m}_{e}_{t}_{e}_{r} _{e}_{s}_{t}_{i}_{m}_{a}_{t}_{e}_{s} _{a}_{n}_{d} residual sumsofsquares for linear versus _{n}_{o}_{n}_{l}_{i}_{n}_{e}_{a}_{r} _{r}_{e}_{g}_{r}_{e}_{s}_{s}_{i}_{o}_{n} _{a}_{n}_{a}_{l}_{y}_{s}_{i}_{s} of simulated Monod dataa
Initial estimates
% Error in:
^{F}^{i}^{n}^{a}^{l} ^{e}^{s}^{t}^{i}^{m}^{a}^{t}^{e}^{s}
RSSj/RSSf'
iMx K,nK Y
>
_{1} 
12.2 
29.8 
14 
2 
_{2} 
_{2}_{9} 
80.4 
37 
9 
_{3} 
_{4}_{1} 
_{1}_{1}_{0} 
56.5 
1 
_{4} 
_{1}_{.}_{2} 
10.4 
8.5 
3 
_{5} 
7.1 
10.8 
5.5 
5 
6 
22.5 
52.4 
20 
1 
_{7} 
7 
12.8 
5.5 
1 
_{8} 
_{1}_{6} 
_{2}_{6}_{.}_{6} 
_{2}_{2} 
_{0} 
_{9} 
_{0} 
_{5}_{.}_{6} 
_{2}_{3}_{.}_{5} 
_{1} 
_{1}_{0} 
_{2}_{1} 
_{3}_{3} 
_{1}_{0} 
_{0}_{.}_{5} 
_{1}_{1} 
_{5}_{.}_{2} 
_{7}_{.}_{2} 
_{2} 
_{2} 
_{1}_{2} 
_{4}_{.}_{8} 
_{4}_{.}_{2} 
_{5} 
_{0} 
K, 
Y 

1.2 
3.5 
71.8 
12.8 
9.5 
337 
1 
2.5 
974 
2.4 
4 
144 
5.2 
6.5 
4.01 
0.4 
2.5 
69.5 
1.2 
1 
15.8 
_{0}_{.}_{2} 
_{0} 
1,150 
_{1}_{.}_{6} 
_{1} 
22,220 
_{1} 
_{0} 
3,430 
_{2}_{.}_{6} 
_{1}_{.}_{5} 
46.1 
_{0} 
_{0} 
75.4 
a True
parameter _{v}_{a}_{l}_{u}_{e}_{s} _{w}_{e}_{r}_{e} _{i}_{d}_{e}_{n}_{t}_{i}_{c}_{a}_{l} for all 12 simulations: eLmax = 0.1, K, = 5, and Y = 0.2.
^{b} Simulations
1 _{t}_{h}_{r}_{o}_{u}_{g}_{h} 6 contain errors having a constant coefficient of variation (0.5%) (relative errors);
simulations 7 through
_{1}_{2} _{c}_{o}_{n}_{t}_{a}_{i}_{n} _{e}_{r}_{r}_{o}_{r}_{s} with a constant standard deviation (0.01) (simple errors).
^{I} _{R}_{a}_{t}_{i}_{o} _{o}_{f} residual _{s}_{u}_{m}_{}_{o}_{f}_{}_{s}_{q}_{u}_{a}_{r}_{e}_{s} _{f}_{o}_{r} initial parameter estimates to residual sumofsquares for final parameter estimates.
_{p}_{o}_{s}_{s}_{i}_{b}_{l}_{e} _{t}_{o} _{u}_{n}_{i}_{q}_{u}_{e}_{l}_{y} estimate parameters for a
_{g}_{i}_{v}_{e}_{n} _{n}_{o}_{n}_{l}_{i}_{n}_{e}_{a}_{r}_{m}_{o}_{d}_{e}_{l} _{w}_{h}_{e}_{n} the sensitivitycoef
ficients predict otherwise.
(particularly for gaseous
_{f}_{r}_{o}_{m} _{s}_{o}_{m}_{e} limitations. It is advantageous since
substrates) but suffers
estimates of growth kinetic parameters may be
_{o}_{b}_{t}_{a}_{i}_{n}_{e}_{d}_{f}_{r}_{o}_{m} _{a} _{s}_{i}_{n}_{g}_{l}_{e} substrate depletioncurve.
_{F}_{u}_{r}_{t}_{h}_{e}_{r}_{,} estimating
_{b}_{y} _{u}_{s}_{i}_{n}_{g} the integrated Monod equation offers an alternative to estimating these parameters for
_{g}_{r}_{o}_{w}_{t}_{h} kinetic parameters
_{g}_{a}_{s}_{e}_{o}_{u}_{s} substrates _{b}_{y} _{u}_{s}_{i}_{n}_{g}
_{i}_{n} _{w}_{h}_{i}_{c}_{h} mass transport
chemostat cultures
limitations _{m}_{a}_{y} _{y}_{i}_{e}_{l}_{d}
unrealistically high
the integrated Monod equation _{c}_{a}_{u}_{s}_{e} the derivative _{f}_{o}_{r}_{m}_{,} from
_{K}_{,} estimates _{(}_{1}_{4}_{)}_{.} The use of
_{i}_{s} limited be
which the inte
_{v}_{e}_{r}_{s}_{i}_{o}_{n} _{i}_{s} _{d}_{e}_{r}_{i}_{v}_{e}_{d}_{,} describes balanced
grated
(steadystate) bacterial growth
describe transient (batch)
_{a}_{n}_{d} _{m}_{a}_{y} fail to
bacterial _{g}_{r}_{o}_{w}_{t}_{h} _{(}_{1}_{3}_{)}_{.}
_{P}_{o}_{w}_{e}_{l}_{l} _{h}_{a}_{s} shown that apparent Monod parame
_{T}_{h}_{e} _{i}_{n}_{t}_{e}_{g}_{r}_{a}_{t}_{e}_{d} Monod equation is advanta
_{g}_{e}_{o}_{u}_{s} _{f}_{o}_{r} _{t}_{h}_{e} estimation of Umax, _{K}_{,}_{,} and Y
A
c
0

O
C
e
c
C
0
dU e 0
U
S1
0o:
B
20C
Time
c 

0 

a 
_{1}_{6} 
Final estimates 

4 

C0 e0 C 
12 
12~~~CV= cvo.o 0.005 

0 

U 

e 
8 
0 _{a}
a,.
_{4} _{.} _{I}_{n}_{i}_{t}_{i}_{a}_{l} estimates ',
O1
.
_{5}
.
10
*i
15
m
20
2S5
Time
3
_{F}_{I}_{G}_{.} _{3}_{.} _{C}_{o}_{m}_{p}_{a}_{r}_{i}_{s}_{o}_{n} of theoretical curves calculat
_{e}_{d} _{b}_{y} _{u}_{s}_{i}_{n}_{g} initial versus final _{(}_{b}_{e}_{s}_{t}_{)} _{,}
K,, and Y
estimates with simulated data points. The data are
_{t}_{h}_{o}_{s}_{e} _{t}_{h}_{a}_{t} _{w}_{e}_{r}_{e} transformed and _{p}_{l}_{o}_{t}_{t}_{e}_{d} in _{F}_{i}_{g}_{.} 2.
cO
I
co
0 am
0
_{0}
_{8}
16
Hour
24
32
40
FIG. 4. Comparison _{o}_{f} theoretical curve with mea 

_{s}_{u}_{r}_{e}_{d} _{H}_{2} concentrations _{(}_{O}_{)} for growth of Desulfovib 

rio sp. 
_{s}_{t}_{r}_{a}_{i}_{n} _{G}_{l}_{l} _{o}_{n} _{H}_{2} _{a}_{s} the sole electron donor. The 

initial 
_{s}_{u}_{b}_{s}_{t}_{r}_{a}_{t}_{e} concentration _{(}_{S}_{O}_{)} _{w}_{a}_{s} estimated 
by 
extrapolating back to t = 0 on _{t}_{h}_{e} _{H}_{2} concentration 
axis.
1458 ROBINSON AND TIEDJE
_{t}_{e}_{r}_{s} determined by using batch growth data may _{d}_{e}_{p}_{e}_{n}_{d} _{o}_{n} _{t}_{h}_{e} _{g}_{r}_{o}_{w}_{t}_{h} rate history of the orga _{n}_{i}_{s}_{m} _{(}_{1}_{3}_{)}_{.} _{L}_{a}_{s}_{t}_{l}_{y}_{,} maintenance requirements, which we neglected for strain Gil, may produce erroneous parameter estimates (particularly of Ks) if ignored. In _{s}_{u}_{m}_{m}_{a}_{r}_{y}_{,} the integrated _{M}_{o}_{n}_{o}_{d} _{e}_{q}_{u}_{a}_{t}_{i}_{o}_{n} _{o}_{f}_{f}_{e}_{r}_{s} an alternative to chemo stat studies forestimating growth kinetic param _{e}_{t}_{e}_{r}_{s}_{,} but it should be used only after its limita _{t}_{i}_{o}_{n}_{s} _{h}_{a}_{v}_{e} been considered.
ACKNOWLEDGMENTS
_{W}_{e} _{t}_{h}_{a}_{n}_{k} _{D}_{a}_{v}_{e} _{M}_{y}_{r}_{o}_{l}_{d} _{f}_{o}_{r} _{u}_{s}_{e} _{o}_{f} _{a} _{c}_{o}_{m}_{p}_{u}_{t}_{e}_{r} program that _{f}_{i}_{t}_{s} _{c}_{u}_{b}_{i}_{c} _{s}_{p}_{l}_{i}_{n}_{e}_{s} _{t}_{o} _{d}_{a}_{t}_{a} and Dan Shelton for many helpful _{d}_{i}_{s}_{c}_{u}_{s}_{s}_{i}_{o}_{n}_{s} _{o}_{n} _{b}_{a}_{c}_{t}_{e}_{r}_{i}_{a}_{l} growth kinetics. We gratefully ac _{k}_{n}_{o}_{w}_{l}_{e}_{d}_{g}_{e} _{t}_{h}_{e} _{e}_{x}_{p}_{e}_{r}_{t} _{t}_{e}_{c}_{h}_{n}_{i}_{c}_{a}_{l} assistance of Walter Smo lenski and thank James Beck for a critical review of the
manuscript. A copy of MONODCRV is available from us
_{u}_{p}_{o}_{n} request.
This work was supported by National Science Foundation _{g}_{r}_{a}_{n}_{t}_{s} DEB 7805321 and DEB 8109994.
LITERATURE CITED
1. 
_{B}_{e}_{c}_{k}_{,} _{J}_{.} _{V}_{.}_{,} _{a}_{n}_{d} K. J. Arnold. 1976. Parameter estimation _{i}_{n} _{e}_{n}_{g}_{i}_{n}_{e}_{e}_{r}_{i}_{n}_{g} _{a}_{n}_{d} _{s}_{c}_{i}_{e}_{n}_{c}_{e}_{,} _{p}_{.} 334350. John Wiley & 
Sons, Inc., New York. 

2. 
Betlach, M. R., and J. M. _{T}_{i}_{e}_{d}_{j}_{e}_{.} 1981. Kinetic explana tion for accumulation of nitrite, nitric oxide, and nitrous _{o}_{x}_{i}_{d}_{e} _{d}_{u}_{r}_{i}_{n}_{g} _{b}_{a}_{c}_{t}_{e}_{r}_{i}_{a}_{l} denitrification. Appl. Environ. Mi crobiol. 42:10741084. 
3. 
_{B}_{e}_{t}_{l}_{a}_{c}_{h}_{,} M. _{R}_{.}_{,} J. M. _{T}_{i}_{e}_{d}_{j}_{e}_{,} and R. B. Firestone. 1981. Assimilatory nitrate uptake in Pseudomonas fluorescens 
studied _{u}_{s}_{i}_{n}_{g} nitrogen13. Arch. Microbiol. 129:135140. 

4. 
_{B}_{u}_{r}_{d}_{e}_{n}_{,} _{R}_{.} _{L}_{.}_{,} _{J}_{.} D. Faires, and A. C. Reynolds. 1978. Numerical _{a}_{n}_{a}_{l}_{y}_{s}_{i}_{s}_{,} _{p}_{.} 116128 and 239245. Prindle, Weber and Schmidt, Boston, Mass. 
_{A}_{P}_{P}_{L}_{.} ENVIRON. MICROBIOL.
5. CornishBowden, A. 1979. Fundamentals of enzyme kinet
_{i}_{c}_{s}_{,} _{p}_{.} 200210. Butterworth, Inc., Boston, Mass.
6. Counotte, G. H. M., and R. A. Prins. 1979. Calculation of _{K}_{m} _{a}_{n}_{d} _{V}_{,}_{,}_{,}_{0}_{,} from substrate concentration versus time _{p}_{l}_{o}_{t}_{.} _{A}_{p}_{p}_{l}_{.} _{E}_{n}_{v}_{i}_{r}_{o}_{n}_{.} Microbiol. 38:758760.
_{7}_{.} _{D}_{r}_{a}_{p}_{e}_{r}_{,} _{N}_{.} _{R}_{.}_{,} and H. Smith. 1981. Applied regression
_{a}_{n}_{a}_{l}_{y}_{s}_{i}_{s}_{,} _{p}_{.} _{4}_{5}_{9}_{.} _{J}_{o}_{h}_{n} Wiley & Sons, Inc., New York. _{8}_{.} _{D}_{u}_{g}_{g}_{l}_{e}_{b}_{y}_{,} _{R}_{.} _{G}_{.}_{,} _{a}_{n}_{d} _{J}_{.} F. Morrison. 1977. The analysis
of _{p}_{r}_{o}_{g}_{r}_{e}_{s}_{s} curves for enzymecatalyzed reactions by nonlinear regression. Biochim. Biophys. Acta 481:297
312.
9. 
_{G}_{r}_{a}_{h}_{a}_{m}_{,} _{J}_{.} _{M}_{.}_{,} and R. P. Canale. 1982. Experimental and 

_{m}_{o}_{d}_{e}_{l}_{i}_{n}_{g} studies of a fourtrophic level predatorprey 

_{s}_{y}_{s}_{t}_{e}_{m}_{.} Microb. Ecol. 8:217232. 

_{1}_{0}_{.} 
_{H}_{a}_{r}_{b}_{a}_{u}_{g}_{h}_{,} _{J}_{.}_{,} and G. BonhamCarter. 1970. Computer 

simulation in geology, p. 6197. John Wiley & Sons, Inc., 

New York. 

11. 
Knowles, G., A. L. Downing, and M. J. Barre"t. 1965. _{D}_{e}_{t}_{e}_{r}_{m}_{i}_{n}_{a}_{t}_{i}_{o}_{n} _{o}_{f} kinetic constants for nitrifying bacteria _{i}_{n} _{m}_{i}_{x}_{e}_{d} _{c}_{u}_{l}_{t}_{u}_{r}_{e}_{,} with the aid of an electronic computer. 

_{J}_{.} Gen. Microbiol. 38:263278. 

_{1}_{2}_{.} 
_{P}_{i}_{r}_{t}_{,} _{S}_{.} _{J}_{.} _{1}_{9}_{7}_{5}_{.} _{P}_{r}_{i}_{n}_{c}_{i}_{p}_{l}_{e}_{s} of microbe and cell cultivation, _{p}_{.} 2228. John _{W}_{i}_{l}_{e}_{y} & _{S}_{o}_{n}_{s}_{,} Inc., New York. 

_{1}_{3}_{.} 
_{P}_{o}_{w}_{e}_{l}_{l}_{,} _{E}_{.} _{0}_{.} _{1}_{9}_{6}_{7}_{.} The growth rate of microorganisms as 

a _{f}_{u}_{n}_{c}_{t}_{i}_{o}_{n} of substrate concentration, p. 3456. In E. 0. 

_{P}_{o}_{w}_{e}_{l}_{l}_{,} C. G. T. Evans, R. E. Strange, and D. W. Tem _{p}_{e}_{s}_{t} _{(}_{e}_{d}_{.}_{)}_{,} _{M}_{i}_{c}_{r}_{o}_{b}_{i}_{a}_{l} _{p}_{h}_{y}_{s}_{i}_{o}_{l}_{o}_{g}_{y} and continuous culture. 

Her Majesty's Stationery Office, London, United King 

dom. 

14. 
_{R}_{o}_{b}_{i}_{n}_{s}_{o}_{n}_{,} J. A., and J. M. Tiedje. 1982. Kinetics of 

_{h}_{y}_{d}_{r}_{o}_{g}_{e}_{n} _{c}_{o}_{n}_{s}_{u}_{m}_{p}_{t}_{i}_{o}_{n} by rumen fluid, anaerobic digestor _{s}_{l}_{u}_{d}_{g}_{e}_{,} and sediment. Appl. Environ. Microbiol. 44:1374 

1384. 

15. 
_{S}_{t}_{r}_{a}_{y}_{e}_{r}_{,} R. 
_{F}_{.}_{,} and _{J}_{.} M. _{T}_{i}_{e}_{d}_{U}_{e}_{.} 1978. Kineticparameters 
of the conversion of methane _{p}_{r}_{e}_{c}_{u}_{r}_{s}_{o}_{r}_{s} to methane in a
_{h}_{y}_{p}_{e}_{r}_{e}_{u}_{t}_{r}_{o}_{p}_{h}_{i}_{c} lake sediment. Appl. Environ. Microbiol.
36:330340.
_{1}_{6}_{.} _{T}_{h}_{o}_{m}_{a}_{s}_{,} G. B., Jr. 1972. Calculus and analytical geome try. p. 7681.
Molto più che documenti.
Scopri tutto ciò che Scribd ha da offrire, inclusi libri e audiolibri dei maggiori editori.
Annulla in qualsiasi momento.