For background information about the quantitative approach to human acid-base physiology,
visit the related site:
https://www.acidbase.org/.
The original text of Peter Stewart's classic book, How to Understand Acid-Base:
A Quantitative Acid-Base Primer for Biology and Medicine, is included in the fully
revised third edition of the book, now entitled Stewart's Textbook of Acid-Base
(Paul Elbers, František Duška and John Kellum, editors, Lulu Press, 2025), now available for purchase at
https://www.acidbase.org/. The third edition
features Stewart's original text plus 20 new fully updated chapters
that highlight advances in the field.
Stewart (1983) introduced a quantitative physicochemical model of acid-base balance
in blood plasma. The Stewart model incorporates three fundamental
physicochemical principles as they apply to a single body fluid compartment (such as
arterial blood plasma) under steady-state conditions: the law of conservation of mass is
always obeyed; electrical neutrality is always maintained; and all statements of chemical
equilibria are simultaneously satisfied. Dissociation equilibria for the carbon
dioxide - bicarbonate - carbonate system are explicitly included. The expression employed
for the carbon dioxide - bicarbonate equilibrium is mathematically equivalent to the
Henderson-Hasselbalch equation. All nonvolatile weak acids
(such as H2PO4-, and plasma proteins) are
characterized by a single equilibrium dissociation constant in Stewart's model.
Figge, Rossing and Fencl (1991) produced electrolyte solutions resembling human serum that contained
albumin as the sole protein moiety. Data collected from these solutions were used in a
least-squares algorithm to develop a more robust quantitative physicochemical model. This
model treats albumin as a polyprotic macromolecule with multiple apparent equilibrium dissociation
constants corresponding to different classes of amino acid side chains (i.e., Arg, Lys,
Asp, Glu, Cys, His, Tyr, amino terminus, carboxyl terminus). The number of side chains in
each class is taken from the known human serum albumin amino acid sequence. The
Figge-Rossing-Fencl model accounts mathematically for two distinct categories of side chains with
respect to their contribution to charge balance. The first category consists of those side
chains with a positively charged acidic form and a neutral conjugate base (i.e., Arg, Lys,
His, and the amino terminus). For example:
-NH3+ ⇄ -NH2 + H+
The second category consists of those side chains with a neutral acidic form and a negatively
charged conjugate base (i.e., Asp, Glu, Cys, Tyr, and carboxyl terminus). For example:
-COOH ⇄ -COO- + H+
As demonstrated in the x-ray crystal structure
of human serum albumin, of the 35 cysteine residues in the protein,
34 form 17 disulfide bridges; hence only one Cys residue is free to ionize.
Apparent equilibrium dissociation constants from the work of Sendroy and Hastings (1927) for the
phosphoric acid - phosphate system ( [ H3PO4 ],
[ H2PO4- ], [ HPO42- ], and
[ PO43- ] ), as applicable to plasma at 38 degrees Celsius,
are explicitly included: pK'1 = 1.915; pK'2 = 6.66;
and pK'3 = 11.78. The Figge-Rossing-Fencl model simultaneously solves
the equilibrium equations governing the following dissociation reactions, and accounts for the net
negative charge contributed by all three ionized species:
H3PO4 ⇄ H2PO4- + H+
H2PO4- ⇄ HPO42- + H+
HPO42- ⇄ PO43- + H+
Within the physiologic pH range, the vast majority of charge attributable to phosphate species derives from
H2PO4- and HPO42-.
The Figge-Rossing-Fencl model is successful in calculating the pH of albumin-containing electrolyte solutions
as well as the pH of filtrands of serum.
Figge, Mydosh and Fencl (1992) further refined the quantitative physicochemical model by incorporating
pKA values for albumin histidine residues as determined by NMR spectroscopy in the study of Labro and
colleagues (1986) and Bos and colleagues (1989). The pKA values are temperature-corrected
to 37 degrees Celsius in the model. This model accounts for the effects of the microenvironments within the
macromolecule of albumin on the pKA values of individual histidine residues. Although the
Figge-Mydosh-Fencl model is successful in many aspects, it does not account for the presence of all 59 lysine
residues in human serum albumin. Furthermore, the Figge-Mydosh-Fencl model does not account for the
neutral-to-base (N–B) structural transition that occurs in human serum albumin between pH 6 and pH 9. This
structural transition features a downward shift in the pKA values of five histidine residues as the
albumin molecule transitions from the N state to the B state. The Figge-Mydosh-Fencl model is limited as it
employs pKA values exclusively from NMR data representing the N state. Hence, the model fails to
account for the B state.
Because of the above limitations, the Figge-Mydosh-Fencl model provides useful results
restricted to the pH range of biologic interest (6.9 to 7.9); outside of this range the model is unreliable.
The model was updated in 2007-2009 and published by Figge (2009) in Stewart's Textbook of Acid-Base, second edition (Chapter 11),
under the title of the Figge-Fencl Quantitative Physicochemical Model of Human Acid-Base Physiology. This model
successfully accounts for all 59 lysine residues in human serum albumin and incorporates information about lysine
residues with unusually low pKA values, in accord with the prior work of Halle and Lindman (1978), and as
suggested by data from tryptophan and tyrosine fluorescence emission spectroscopy studies by Dockal and colleagues
(2000). As in the Figge-Mydosh-Fencl (1992) model, pK(a) values for 13 of 16 albumin histidine residues in the
Figge-Fencl model are based on NMR spectroscopy data (temperature-corrected from 25 to 37 degrees Celsius).
The model also accounts for the neutral-to-base (N-B) structural transition of human serum albumin over the pH range
of 6 to 9 (see below). The titration curve of human serum albumin at 37 degrees Celsius as predicted by the Figge-Fencl model
closely tracks with the experimental data points of Niels Fogh-Andersen and colleagues (1993) over the pH range of
5 to 9.
Human serum albumin undergoes several structural transitions as a function of pH.
The N-B (neutral-to-base) transition occurs between pH 6 and 9, which is important because this interval includes the physiologic pH range.
Human serum albumin is organized into three structurally homologous domains, denoted as HSA domain I, HSA domain II and HSA domain III [Dockal, Carter and Rüker (1999)].
A variety of spectroscopic methods including far-UV circular dichroism (CD), near-UV CD, and fluorescence emission (from tryptophan and tyrosine residues)
have been employed to study the structural transitions of human albumin and the three recombinant domains of albumin [Dockal, Carter and Rüker (2000)].
The role of histidine residues in the N-B transition has been studied by NMR spectroscopy [Labro and Janssen (1986); Bos et al. (1989)].
The fluorescence spectrum of bovine serum albumin has also been studied in the pH range of 3 to 10 [Halfman and Nishida (1971)].
The N-B transition has been described in terms of a two-state model [Janssen, Van Wilgenburg and Wilting (1981)].
The N state exists at lower pH values, and the B state at higher pH values.
This interpretation is supported by the far-UV CD data which demonstrate a slight reduction in alpha-helical content of albumin as the pH ranges from
7.4 to 9.0 [Dockal, Carter and Rüker (2000)].
Furthermore, near-UV CD data demonstrate that HSA domains I and II undergo a tertiary structural isomerization
in the pH range of the N-B transition [Dockal, Carter and Rüker (2000)].
HSA domain III is not involved in the N-B structural transition.
Based on model calculations, there are most likely five conformation-linked histidine residues that undergo a downward pK shift as albumin
transitions from the N to the B conformation [Bos et al. (1989)].
The five histidine residues that participate in the N-B transition have been assigned to HSA domain I
based on an analysis of NMR data [Bos et al. (1989)].
Calcium ions can exert a strong influence on the N-B transition. Calcium ions induce a downward shift in the pK(a) of several histidine residues at constant pH
and a concomitant downward shift in the midpoint pH of the N-B transition [Labro and Janssen (1986); Bos et al. (1989); Janssen, Van Wilgenburg and Wilting (1981)].
Consequently, the addition of calcium results in the release of protons and a shift from the N to the B conformation.
The Figge-Fencl quantitative physicochemical model of acid-base physiology (version 3.0), presented below,
incorporates an empiric function that models the N-B transition by downshifting the pK(a) values of five histidine residues
located within HSA domain I.
The magnitude of the pK downshift is 0.4 units, as estimated by Labro and Janssen (1986).
In the presence of physiologic calcium ion concentrations (2.5 mM), the midpoint of the pH range for the N-B transition is
approximately 6.9 [Janssen, Van Wilgenburg and Wilting (1981)].
The Figge-Fencl model was updated in 2012, and the most recent version is 3.0, which is
now featured on https://www.acid-base.org/. The Figge-Fencl model
version 3.0 replicates the key results of the Figge-Mydosh-Fencl (1992) model within the pH range of
biologic interest (6.9 to 7.9), while at the same time incorporating the contribution of all 59 lysine
residues.
The model is also described in the appendix of Figge, Bellomo and Egi (2018) and in Chapter 11 of Stewart's Textbook of Acid-Base
(Paul Elbers, František Duška and John Kellum, editors, Third Edition, Lulu Press, 2025).
Version 3.0 incorporates key enhancements from earlier models, and features an improved least squares fit to the
original data of Figge, Rossing and Fencl (1991) compared with the Figge-Mydosh-Fencl (1992) model and the
Figge-Fencl model of 2009. Version 3.0 also improves the performance of the model down to pH 4, extending
the useful range from pH 4 to 9. The titration curve of human serum albumin at 37 degrees Celsius as predicted
by the Figge-Fencl model version 3.0 closely tracks with the experimental data points
of Niels Fogh-Andersen and colleagues (1993) over the pH range of 4 to 9. The Figge-Fencl model
version 3.0 gives results equivalent to those of the Figge-Mydosh-Fencl model within the pH range of
biologic interest (6.9 to 7.9). Technical details about model version 3.0 can be accessed through the links and text below.
Key Features of the model include:
● The expression employed for the carbon dioxide - bicarbonate equilibrium is mathematically equivalent to the
Henderson-Hasselbalch equation.
● Apparent equilibrium dissociation constants from the work of Sendroy and Hastings (1927) for the
phosphoric acid - phosphate system ( [ H3PO4 ],
[ H2PO4- ], [ HPO42- ], and
[ PO43- ] ), as applicable to plasma at 38 degrees Celsius,
are explicitly included.
● The model treats albumin as a polyprotic macromolecule with multiple apparent equilibrium dissociation
constants corresponding to different classes of amino acid side chains (i.e., Arg, Lys,
Asp, Glu, Cys, His, Tyr, amino terminus, carboxyl terminus). The number of side chains in
each class is taken from the known human serum albumin amino acid sequence.
● The model incorporates pKA values for albumin histidine residues as determined by NMR spectroscopy
in the study of Bos and colleagues (1989). The pKA values are temperature-corrected to 37 degrees Celsius
in the model. This model accounts for the effects of the microenvironments within the macromolecule of albumin on
the pKA values of individual histidine residues.
● The model successfully accounts for the contribution of all 59 lysine residues in human serum albumin and
incorporates information about lysine residues with unusually low apparent pKA values, in accord with the
prior work of Halle and Lindman (1978), and as suggested by data from tryptophan and tyrosine fluorescence emission
spectroscopy studies by Dockal and colleagues (2000).
● The model accounts for the neutral-to-base (N-B) structural transition of human serum albumin over the pH range
of 6 to 9.
● The model accounts for the anomalously low average pK(a) value of glutamic and aspartic acid residues in
albumin. This feature allows the model to provide useful functionality down to a pH of approximately 4.0.
● The model is backwards-compatible with the Figge-Mydosh-Fencl (1992)
model within the pH range of 6.9 to 7.9, thereby preserving consistency with prior results.
Abstract.
The Figge-Fencl quantitative physicochemical model of human acid-base physiology in blood
plasma (version 3.0) is presented below. Following Stewart (1983), the model incorporates three fundamental
physicochemical principles as they apply to a single body fluid compartment (such as
arterial blood plasma) under steady-state conditions: the law of conservation of mass is
always obeyed; electrical neutrality is always maintained; and all statements of chemical
equilibria are simultaneously satisfied.
Dissociation equilibria for the carbon dioxide - bicarbonate - carbonate system
are explicitly included. The expression employed for the carbon dioxide - bicarbonate
equilibrium is mathematically equivalent to the Henderson-Hasselbalch equation. Hence,
the Henderson-Hasselbalch equation is always satisfied in the Figge-Fencl model (version 3.0). This is
a necessary but not sufficient condition to describe the acid-base status of a given body
fluid compartment.
Dissociation equilibria for major species of weak acids and their conjugate bases are
incorporated into the Figge-Fencl model (version 3.0). These include the phosphoric acid - phosphate
system, and dissociable amino acid side chains of
albumin. Plasma globulins play only a minor role in acid-base balance and are not
explicitly included in the current version of the Figge-Fencl model.
Human serum albumin is treated as a polyprotic macromolecule with
multiple equilibrium dissociation constants corresponding to different classes of amino
acid side chains. As in the Figge-Mydosh-Fencl (1992) model, pK(a) values for 13 of 16 albumin
histidine residues in the Figge-Fencl model (version 3.0) are based on NMR spectroscopy data
(temperature-corrected from 25 to 37 degrees Celsius). The Figge-Fencl model (version 3.0) features three
major advances over the Figge-Mydosh-Fencl (1992) model.
First, the Figge-Fencl model (version 3.0) takes into account the neutral-to-base (N-B) structural
transition of human serum albumin over the pH range of 6 to 9. Second, the Figge-Fencl model (version 3.0) explicitly
accounts for the contribution of all 59 lysine residues in albumin. The model features a
small number of lysine residues with unusually low pK(a) values, the existence of which is
suggested by spectroscopic and x-ray crystallographic data. Third, the Figge-Fencl model (version 3.0) accounts for the
anomalously low average pK(a) value of glutamic and aspartic acid residues in albumin. This feature allows the model to
provide useful functionality down to a pH of approximately 4.0.
The Figge-Fencl model (version 3.0) gives results equivalent to those of the Figge-Mydosh-Fencl (1992) model within the pH range of
biologic interest (6.9 to 7.9). The Figge-Fencl model (version 3.0) has been optimized against the data of Figge, Rossing and
Fencl (1991) for albumin-containing solutions using a least-squares algorithm. The Figge-Fencl model (version 3.0) demonstrates an
improved least-squares fit to these data compared with the Figge-Mydosh-Fencl model. In addition, the albumin titration curve
that is generated by the Figge-Fencl model (version 3.0) closely tracks with the experimental data points of Fogh-Andersen,
Bjerrum, and Siggaard-Andersen (1993) within the pH range of 4 to 9, providing an additional level of validation. Hence, the model
allows for robust analysis of acid-base phenomena over the pH window of 4 to 9.
At pH 7.40, the Figge-Fencl model (version 3.0) predicts that the charge contributed by 4.4 g / dL of albumin is approximately
-12.3 mEq / L. The Figge-Fencl model (version 3.0) predicts that the titration curve of human serum albumin (charge displayed
by albumin versus pH) is approximately linear over the pH range of biologic interest (6.9 to 7.9). Within this pH interval,
the Figge-Fencl model (version 3.0) predicts that the approximate charge contributed by albumin in human plasma at a
given pH is:
[ Albx- ] = -10 x [ Albumin ] x ( 0.123 x pH - 0.631 );
where [ Albx- ] is in mEq/L and [ Albumin ] is in g/dL. This result exactly replicates the result
obtained for the Figge-Mydosh-Fencl model (1992) over the same pH range.
The molar buffer capacity of albumin is calculated by taking the additive inverse of the first derivative
of [ Albx- ] with respect to pH:
- d [ Albx- ] / d pH
At pH = 7.40, the molar buffer capacity of albumin is
8.2 Eq / mol / pH. Thus, for [ Albumin ]
of 4.4 g/dL, the buffer capacity of albumin is: 5.4 mEq/L/pH. The buffer
capacity of albumin is used in the formulation of the classic van Slyke equation for
plasma. The approximate linearity of the albumin titration curve over the pH range of biologic interest is
critical for the successful derivation of the van Slyke equation for plasma.
The Figge-Fencl quantitative physicochemical model (version 3.0) can be used in conjunction
with a computer application to illustrate the effects of independent variables on
acid-base status in plasma-like solutions containing albumin. The application program
will solve the following function for pH:
pH = fpH { SID, PCO2, [ Pitot ], [ Albumin ] }
Introductory Notes.
[1] Peter A. Stewart introduced a quantitative physicochemical model of acid-base balance
in blood plasma.
Stewart incorporated the principle of electrical neutrality and
accounted for the electrical charges contributed by all ionized species. Dissociation
equilibria for the carbon dioxide - bicarbonate - carbonate system were explicitly
included. The expression employed by Stewart for the carbon dioxide - bicarbonate
equilibrium is mathematically equivalent to the Henderson-Hasselbalch equation. All
nonvolatile weak acids (H2PO4-, and plasma proteins)
were characterized by a single
equilibrium dissociation constant in Stewart's model.
[2] Figge, Rossing and Fencl
produced electrolyte solutions resembling human serum that contained
albumin as the sole protein moiety. Data collected from these solutions were used in a
least-squares algorithm to develop a more robust quantitative physicochemical model. This
model treated albumin as a polyprotic macromolecule with multiple equilibrium dissociation
constants corresponding to different classes of amino acid side chains (i.e., Arg + Lys,
Asp + Glu, Cys, His, Tyr, amino terminus, carboxyl terminus). This model was successful in
calculating the pH of albumin-containing electrolyte solutions as well as the pH of
filtrands of serum.
[3] Figge, Mydosh and Fencl
further refined the quantitative physicochemical model by incorporating
pK(a) values for albumin histidine residues as determined by NMR spectroscopy. The pK(a)
values were temperature-corrected to 37 degrees Celsius in the model. Based on a model
compound, a correction factor of -0.27 was used to adjust the histidine pK(a) values from 25 degrees Celsius (298 K)
to 37 degrees Celsius (310 K):
pK(a)310 = pK(a)298 - 0.27
This model accounted for the effects of the microenvironments within
the macromolecule of albumin on the pK(a) values of individual histidine residues.
The same temperature-corrected values are employed in the Figge-Fencl quantitative physicochemical
model of acid-base physiology (version 3.0), presented below.
The imidazole groups of histidine residues are of particular importance because they can
titrate within the pH range of biologic interest. Although the Figge-Mydosh-Fencl model was successful in many
aspects, it did not account for the presence of all 59 lysine residues in human serum albumin.
[4] Human serum albumin undergoes several structural transitions as a function of pH. The
N-B (neutral-to-base) transition occurs between pH 6 and 9, which is important because this
includes the physiologic pH range. Human serum albumin is organized into three
structurally homologous domains, denoted 1, 2 and 3. A
variety of spectroscopic methods including far-UV circular dichroism (CD), near-UV CD, and
fluorescence emission (from tryptophan and tyrosine residues) have been employed to study
the structural transitions of human albumin and the three recombinant domains of albumin. The role of
histidine residues in the N-B transition has been studied by NMR spectroscopy. The
fluorescence spectrum of bovine serum albumin has also been studied in the pH range of 3 to 10.
The N-B transition has been described in terms of a two-state model.
The N state exists at lower pH values, and the B state at high pH.
This interpretation is supported by the far-UV CD data which demonstrate a slight reduction
in alpha-helical content of albumin as the pH ranges from 7.4 to 9.0. Furthermore,
near-UV CD data demonstrate that albumin domains 1 and 2 undergo a tertiary structural
isomerization in the pH range of the N-B transition. Domain 3 is
not involved in the N-B structural transition. Based on model calculations, there are
most likely five conformation-linked histidine residues that undergo a downward pK shift
as albumin transitions from the N to the B conformation. The five
histidine residues that participate in the N-B transition have been assigned to domain 1 of
human serum albumin based on an analysis of NMR data. Calcium
ions can exert a strong influence on the N-B transition.
Calcium ions induce a downward shift in the pK(a) of several histidine residues at constant
pH and a concomitant downward shift in the midpoint pH of the N-B transition.
Consequently, the addition of calcium results in the release of protons
and a shift from the N to the B conformation.
[5] The Figge-Fencl quantitative physicochemical model of acid-base physiology (version 3.0), presented
below, incorporates an empiric function that models the N-B transition by downshifting the
pK(a) values of five histidine residues located within domain 1 of human serum albumin.
The magnitude of the pK downshift is 0.4 units. In the presence of physiologic calcium ion concentrations (2.5 mM), the
midpoint of the pH range for the N-B transition is approximately 6.9.
[6] Buried lysine residues with unusually low pK(a) values (e.g., pK(a) of 5.7) have been
documented in the literature. The
available evidence suggests that there are potentially six lysine residues in human
serum albumin that have an unusually low pK(a). This interpretation is supported by data
from tryptophan and tyrosine fluorescence emission studies. Human serum
albumin contains a single tryptophan residue at position 214, which is located within
domain 2. Tryptophan fluorescence can be excited by a wavelength of 295 nm, which does not
excite tyrosine residues. When tryptophan-specific fluorescence is studied in the context
of intact human serum albumin, there is a decrease in fluorescence intensity as the pH is
raised from 7.4 to 9.0. When tyrosine
fluorescence is studied from recombinant domain 3, there is a decrease in fluorescence
intensity as the pH is raised from 6 to 9.
The decrease of the tyrosine fluorescence signal intensity could be partially explained by
deprotonation of the phenolic hydroxyl group of some tyrosine side chains. Furthermore,
deprotonated epsilon-amino groups of lysine side chains are known to quench the
fluorescence signal from both tryptophan and tyrosine residues. Therefore, deprotonated
lysine residues in close proximity to tryptophan and tyrosine are candidate quenching
groups. Based on the known x-ray crystal structure
of human albumin [ Human Serum Albumin Crystal Structure ],
the epsilon-amino groups of lysine-525, 414, 432, and 534 are within five Angstroms of
tyrosine-401, 411, 452, and 497, respectively. These tyrosine residues are all located
within domain 3 of human serum albumin. Structural rearrangements
associated with the N-B transition (which do not affect domain 3) are not likely to be a
factor in the decrease of the fluorescence signal from these particular tyrosine residues.
This suggests that the adjacent lysine epsilon-amino groups could be at least partially
responsible for quenching the tyrosine fluorescence signal as they deprotonate within the
pH range of 6 to 9. Likewise, lysine-199 and lysine-195 are positioned 3.7 and 7.4
Angstroms, respectively, from tryptophan-214. Thus, the decrease in the fluorescence signal
from tryptophan-214 might be due to changes in secondary or tertiary structure of domain 2,
and/or quenching from one or both adjacent lysine epsilon-amino groups as they deprotonate
within the slightly alkaline pH range. These data suggest that there are potentially six
lysine epsilon-amino groups in human albumin that exhibit an anomalously low pK(a) value.
Similar fluorescence data were presented regarding bovine serum albumin.
[7] The exact number of low-titrating lysine residues in human serum albumin is not known. Trial pK(a) values for the
low-titrating lysine residues in the Figge-Fencl model (version 3.0) are estimated from the albumin titration curve of Halle and Lindman
(corrected from 22 to 37 degrees C). Based on this analysis, there are nine low-titrating lysine residues
distributed in two clusters. One smaller cluster features pK(a) values near 6. The second, larger cluster, features pK(a)
values near 8.
The number and exact distribution of low-titrating lysine residues are optimized in the Figge-Fencl model (version 3.0) using an
optimization algorithm against the data of Figge, Rossing and Fencl for albumin-containing solutions at 37 degrees C.
The computer program incorporates three groups of trial pK(a) values for low-titrating lysine residue. There are 21 sets of trial pK(a) values within each group.
The group number is designated in the computer program by the parameter, i5, which assumes vaues of 0 to 2 (inclusive).
The set number is designated in the computer program by the parameter, i, which assumes values of 0 to 200 (inclusive)
in increments of 10. For group 1 (i5=0), the 21 trial sets are:
Group 1, Set 01 (i=000): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.500; LYS4 = 7.675; LYS5 = 7.850; LYS6 = 8.025};
Group 1, Set 02 (i=010): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.510; LYS4 = 7.685; LYS5 = 7.860; LYS6 = 8.035};
Group 1, Set 03 (i=020): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.520; LYS4 = 7.695; LYS5 = 7.870; LYS6 = 8.045};
. . .
Group 1, Set 21 (i=200): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.700; LYS4 = 7.875; LYS5 = 8.050; LYS6 = 8.225}.
The number of lysine residues with a pK(a) of LYS1
is designated as N1; the number with a pK(a) of LYS2 is designated as N2; etc. Each parameter, N1 through N6, is allowed to assume a value of
0, 1, or 2. All possible combinations of values are tested for inclusion in the model. Up to nine low-titrating lysine residues
are allowed in the model, so that (N1 + N2 + N3 + N4 + N5 + N6) < 10. These constraints result in 701 possible combinations for N1 through
N6.
The entire process is then repeated a second time for group 2 (i5=1):
Group 2, Set 01 (i=000): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.500; LYS4 = 7.700; LYS5 = 7.900; LYS6 = 8.100};
Group 2, Set 02 (i=010): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.510; LYS4 = 7.710; LYS5 = 7.910; LYS6 = 8.110};
Group 2, Set 03 (i=020): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.520; LYS4 = 7.720; LYS5 = 7.920; LYS6 = 8.120};
. . .
Group 2, Set 21 (i=200): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.700; LYS4 = 7.900; LYS5 = 8.100; LYS6 = 8.300}.
And finally a third time for group 3 (i5=2):
Group 3, Set 01 (i=000): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.500; LYS4 = 7.725; LYS5 = 7.950; LYS6 = 8.175};
Group 3, Set 02 (i=010): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.510; LYS4 = 7.735; LYS5 = 7.960; LYS6 = 8.185};
Group 3, Set 03 (i=020): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.520; LYS4 = 7.745; LYS5 = 7.970; LYS6 = 8.195};
. . .
Group 3, Set 21 (i=200): {LYS1 = 5.8; LYS2 = 6.15; LYS3 = 7.700; LYS4 = 7.925; LYS5 = 8.150; LYS6 = 8.375}.
The pK(a) for normally-titrating lysine residues in the model,
LYS7, is assigned trial values of 10.3; 10.4; 10.5; . . .; 10.8 by the computer program, in
accordance with standard textbook values. The value of LYS7 is calculated in the program from the parameter, L7, which
assumes values of 103 to 108 (inclusive).
This model accounts for the contribution of all 59 lysine residues in human serum albumin.
Hence, the number of normally-titrating lysine residues, N7, is given by: N7 = 59 - N1 - N2 - N3 - N4 - N5 - N6.
Arginine residues are considered separately from lysine. Each of the 24 arginine resides in albumin is assigned
a pK(a) of 12.5, in accordance with standard textbook values.
[8] In the Figge-Fencl quantitative model (version 3.0), presented below, each of the 18 tyrosine residues
in albumin is assigned a pK(a) value of 11.7, in accordance with a spectrophotometrically
determined value. The formal possibility of anomalously low pK(a) values
for a subset of tyrosine residues is not addressed in this model.
[9] In the Figge-Fencl quantitative model (version 3.0), presented below, the 36 aspartic acid and 62
glutamic acid residues in albumin are each assigned a pK(a) of 3.9. This is in recognition of the
anomalously low average pK(a) value of glutamic and aspartic acid residues in albumin. Cysteine is assigned a
pK(a) of 8.5, the amino terminus is assigned a pK(a) of 8.0, and the carboxyl terminus is
assigned a pK(a) of 3.1. These are consistent with standard textbook values
and were previously employed in Figge, Mydosh and Fencl.
[10] An incompletely characterized histidine residue has a temperature-corrected pK(a) < 5.2 and is assigned a value of
HIS14 = 5.10 in the Figge-Fencl model (version 3.0). A second incompletely characterized histidine residue has a temperature-corrected
pK(a) in the range of 6.7 to 7.7 (inclusive). The computer program assigns trial pK(a) values for HIS15 of 6.7; 6.8; 6.9; . . .; 7.7.
The value of HIS15 is calculated from the parameter, N0, which assumes values of 67 to 77 (inclusive) in the computer program.
The pK(a) value of one histidine residue that could not be determined by NMR is
designated as HIS16 and is arbitrarily assigned a value of 6.2 in the model, in keeping with standard textbook values.
[11] A detailed model for the contribution of plasma globulins is difficult to develop due
to the marked heterogeneity of plasma globulin species. Based on liquid phase preparative
isoelectric focusing of native human immunoglobulin molecules, the distribution of
isoelectric points for IgG is 4.35 to 9.95, with a dominant peak between pH 7 and 9.95,
centered at pH 8.2. Thus a significant fraction of IgG molecules will carry a positive
charge within the physiologic pH range. Hence, at physiologic pH values, it is expected
that the positive charges contributed by IgG molecules will at least partially offset the
negative charges carried by alpha- and beta-globulin fractions as well as IgA, the
majority of IgM and the remainder of IgG molecules.
[12] K1, K2 and K3 are the apparent equilibrium dissociation constants for phosphoric
acid for plasma. pK1 =
1.915; pK2 = 6.66; pK3 = 11.78.
[13] The constant Kc1, governing the carbon dioxide - bicarbonate equilibrium, is derived
directly from parameters in the Henderson-Hasselbalch equation. The solubility of
CO2 in plasma is: 0.230 mmol / L / kPa x 0.13332236842105 kPa / Torr =
0.0307 mmol / L / Torr.
Kc1 = (10-6.1) (0.0307) / (1000) = 2.44 x 10-11.
Hence, calculations using Kc1 in the Figge-Fencl model (version 3.0) yield results identical to those
calculated with the Henderson-Hasselbalch equation.
[14] The constant Kc2, the second dissociation constant for carbonic acid, is calculated
from the formula (equation 9) given by Harned and Scholes. At 37 degrees
Celsius (310 K), the formula yields: log Kc2 = (-2902.39 / 310) + 6.4980 - (0.02379 x 310)
= -10.239. Hence, at zero ionic strength, pKc2 = 10.239. The correction factor for an
ionic strength of 0.15 M is approximately 0.022. Hence pKc2 = 10.261, and kc2 = 5.5E-11.
Due to the fact that carbonate ion has a charge of -2, Kc2 when expressed in Eq / L is
1.1E-10.
[15] The computer program is written in BASIC, using double-precision floating point
arithmetic. This program optimizes the values of i5, i, L7, and N0 through N7 using an optimization algorithm against the data of Figge,
Rossing and Fencl. The algorithm finds the subset of all solutions that satisfy the following criteria: (a) the absolute value of the mean
deviation, { Σ(calculated pH - measured pH) } / 65, is 0.0035 or less;
and, (b) the slope of the regression line (calculated pH versus measured pH) is 1.0000 +/- 0.0068. From that subset, the single
solution that minimizes the sum of
squares of deviations is selected. The program analyzes a total of 2,914,758 combinations of values for i5, i, L7, N0, and N1 through N6
(3 x 21 x 6 x 11 x 701 = 2,914,758). The program runs in about 35 hours on an Intel(R) Pentium(R) 2.13 GHz processor. It performs approximately
2.3 x 1012 (2.3 trillion) double-precision floating point calculations. This large number arises from the fact that an iterative
procedure is required to solve the model equation for each combination of variables, and this is repeated 65 times to accommodate each data
point.
[16] A model summary and a simple formula for calculating the
predicted charge displayed by albumin over the pH range of 6.9 to 7.9 are also presented. Backwards-compatibility with the Figge-Mydosh-Fencl
model is one of the design goals of this project to preserve consistency with prior results. Hence the performace of the new model is expected to closely
approximate the performance of the Figge-Mydosh-Fencl model over the pH range of 6.90 to 7.90. This requirement is tested with separate code using linear regression
analysis of data points 6.90 to 7.90 at 0.01 increments (N=101). The slope is required to be 0.123 and intercept -0.631, to relplicate the result from the
Figge-Mydosh-Fencl model to 3 decimal points. Results of this analysis are given below. This requirement could be added to the selection criteria in the code below,
but this would require extra compute time.
The molar buffer capacity
of albumin is calculated by taking the additive inverse of the first derivative of [ Albx- ]
with respect to pH: - d [ Albx- ] / d pH. The first derivative is evaluated at pH = 7.40.
Following this, a
formal mathematical representation of the optimized model is presented.
Computer Source Code.
NOTE: this code is for educational use only. It is not to be used for clinical purposes.
Sub Model()
Rem: Figge-Fencl Quantitative Physicochemical Model
Rem: of Human Acid-Base Physiology (Version 3.0).
Rem:
Rem: Program by James J. Figge, MD, MBA, FACP.
Rem: Copyright 2003 - 2026 James J. Figge. Update published 28 April, 2013;
Rem: Update of computer source code published 27 October, 2013; and 25 January, 2026
Rem: on https://www.acid-base.org.
Close #1
Dim pHm(65), SID(65), PCO2(65), Pi(65), Alb(65)
rownum = 1
colnum = 1
rownum = ActiveCell.Row
colnum = ActiveCell.Column
Worksheets("Sheet1").Activate
sum1 = 0
sum2 = 0
sum3 = 0
sum4 = 0
sum5 = 0
For rownum = 1 To 65
pHm(rownum) = ActiveSheet.Cells(rownum, 2)
SID(rownum) = ActiveSheet.Cells(rownum, 3)
PCO2(rownum) = ActiveSheet.Cells(rownum, 4)
Pi(rownum) = ActiveSheet.Cells(rownum, 5)
Alb(rownum) = ActiveSheet.Cells(rownum, 6)
sum1 = sum1 + pHm(rownum)
sum2 = sum2 + SID(rownum)
sum3 = sum3 + PCO2(rownum)
sum4 = sum4 + Pi(rownum)
sum5 = sum5 + Alb(rownum)
Next rownum
Rem: Kc1 is derived from the parameters in the Henderson-Hasselbalch
Rem: equation. pK = 6.1; a = 0.230 mM / kPa; 1 Torr = 0.13332236842105 kPa
Rem: The value of Kc1 is 2.44E-11 (Eq / L)^2 / Torr.
Rem: Kc2 is calculated from Harned and Scholes (1941) for 37 degrees C and ionic
Rem: strength 0.15 M. The value of Kc2 is 5.5E-11 mol / L x 2 = 1.1E-10 Eq / L.
Rem: K1, K2, and K3 for the phosphoric acid - phosphate system are from Sendroy and
Rem: Hastings (1927).
Const kw = 0.000000000000044
Const Kc1 = 0.0000000000244
Const Kc2 = 0.00000000011
Const K1 = 0.0122
Const K2 = 0.000000219
Const K3 = 0.00000000000166
Const LYS1 = 5.8
Const LYS2 = 6.15
Const L3 = 7500
Const L4 = 7675
Const L5 = 7850
Const L6 = 8025
Rem: HIS 14 has a pK of less than 5.2; the pK value is set at 5.1
Rem: HIS 15 has a pK in the range of 6.7 to 7.7 (inclusive)
Rem: HIS 16 has an unknown pK; the pK value is arbitrarily set at 6.2
Const HIS14 = 5.1
Const HIS16 = 6.2
minss = 9999999
For i5 = 0 To 2
For i = 0 To 200 Step 10
LYS3 = (L3 + i) / 1000
LYS4 = (L4 + 25 * i5 + i) / 1000
LYS5 = (L5 + 50 * i5 + i) / 1000
LYS6 = (L6 + 75 * i5 + i) / 1000
For L7 = 103 To 108
LYS7 = L7 / 10
For N0 = 67 To 77
HIS15 = N0 / 10
For N1 = 0 To 2
For N2 = 0 To 2
For N3 = 0 To 2
For N4 = 0 To 2
For N5 = 0 To 2
For N6 = 0 To 2
If N1 + N2 + N3 + N4 + N5 + N6 > 9 Then GoTo getnext
N7 = 59 - N1 - N2 - N3 - N4 - N5 - N6
ss = 0
s = 0
abvs = 0
sx = 0
sxx = 0
sy = 0
syy = 0
sxy = 0
For j = 1 To 65
High = 14
Low = 1
calculatepH:
pH = (High + Low) / 2
Rem: H is hydrogen ion activity (also used as an approximation of [H+])
H = 10 ^ -pH
HCO3 = Kc1 * PCO2(j) / H
CO3 = Kc2 * HCO3 / H
FNX = K1 * H * H + 2 * K1 * K2 * H + 3 * K1 * K2 * K3
FNY = H * H * H + K1 * H * H + K1 * K2 * H + K1 * K2 * K3
FNZ = FNX / FNY
P = Pi(j) * FNZ
Netcharge = SID(j) + 1000 * (H - kw / H - HCO3 - CO3) - P
Rem: NB accounts for histidine pK shift due to the NB transition
NB = 0.4 * (1 - (1 / (1 + (10 ^ (pH - 6.9)))))
Rem: Calculate charge on albumin
Rem: alb2 accumulates results
Rem: cysteine residue
alb2 = -1 / (1 + 10 ^ (-(pH - 8.5)))
Rem: glutamic acid and aspartic acid residues
alb2 = alb2 - 98 / (1 + 10 ^ (-(pH - 3.9)))
Rem: tyrosine residues
alb2 = alb2 - 18 / (1 + 10 ^ (-(pH - 11.7)))
Rem: arginine residues
alb2 = alb2 + 24 / (1 + 10 ^ (pH - 12.5))
Rem: lysine residues
alb2 = alb2 + N1 / (1 + 10 ^ (pH - LYS1))
alb2 = alb2 + N2 / (1 + 10 ^ (pH - LYS2))
alb2 = alb2 + N3 / (1 + 10 ^ (pH - LYS3))
alb2 = alb2 + N4 / (1 + 10 ^ (pH - LYS4))
alb2 = alb2 + N5 / (1 + 10 ^ (pH - LYS5))
alb2 = alb2 + N6 / (1 + 10 ^ (pH - LYS6))
alb2 = alb2 + N7 / (1 + 10 ^ (pH - LYS7))
Rem: 16 different histidine residues
Rem: correction factor to convert HIS pK(a) from 25 deg C to 37 deg C is approx -0.27
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 7.12 + NB))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 7.22 + NB))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 7.1 + NB))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 7.49 + NB))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 7.01 + NB))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 7.31))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 6.75))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 6.36))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 4.85))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 5.76))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 6.17))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 6.73))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 5.82))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - HIS14))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - HIS15))
alb2 = alb2 + 1 / (1 + 10 ^ (pH - HIS16))
Rem: amino terminus
alb2 = alb2 + 1 / (1 + 10 ^ (pH - 8))
Rem: carboxyl terminus
alb2 = alb2 - 1 / (1 + 10 ^ (-(pH - 3.1)))
alb2 = alb2 * 1000 * 10 * Alb(j) / 66500
Netcharge = Netcharge + alb2
If Abs(Netcharge) < 0.0000001 Then GoTo complete
If Netcharge < 0 Then High = pH
If Netcharge > 0 Then Low = pH
GoTo calculatepH
complete:
ss = ss + (pHm(j) - pH) * (pHm(j) - pH)
s = s + (pHm(j) - pH)
abvs = abvs + Abs(pHm(j) - pH)
sx = sx + pHm(j)
sxx = sxx + pHm(j) * pHm(j)
sy = sy + pH
syy = syy + pH * pH
sxy = sxy + pHm(j) * pH
Next j
If ss > minss Then GoTo getnext
n = 65
If Abs(10000 * s / n) > 35 Then GoTo getnext
Slope = (n * sxy - sx * sy) / (n * sxx - sx * sx)
If Abs(10000 * Slope - 10000) > 68 Then GoTo getnext
minss = ss
Open "model-results" For Output As #1
Print #1, "Checksum1 =", sum1
Print #1, "Checksum2 =", sum2
Print #1, "Checksum3 =", sum3
Print #1, "Checksum4 =", sum4
Print #1, "Checksum5 =", sum5
Print #1, " "
Print #1, "abvs / n =", abvs / 65, "s= ", s, "ss= ", ss
Print #1, " "
Print #1, "HIS14 =", HIS14, "HIS15 = ", HIS15, "HIS16 = ", HIS16
Print #1, " "
Print #1, "LYS1 = ", LYS1, "; N1 = ", N1
Print #1, "LYS2 = ", LYS2, "; N2 = ", N2
Print #1, "LYS3 = ", LYS3, "; N3 = ", N3
Print #1, "LYS4 = ", LYS4, "; N4 = ", N4
Print #1, "LYS5 = ", LYS5, "; N5 = ", N5
Print #1, "LYS6 = ", LYS6, "; N6 = ", N6
Print #1, "LYS7 = ", LYS7, "; N7 = ", N7
n = 65
Slope = (n * sxy - sx * sy) / (n * sxx - sx * sx)
incpt = (sy * sxx - sx * sxy) / (n * sxx - sx * sx)
vincpt = sy / n - Slope * sx / n
r = (n * sxy - sx * sy) / Sqr(n * sxx - sx * sx) / Sqr(n * syy - sy * sy)
Var = (syy - incpt * sy - Slope * sxy) / (n - 2)
varslope = n * Var / (n * sxx - sx * sx)
stndevslope = Sqr(varslope)
Rem: t(n-2, alpha/2) for n=65 is 2.3870, where alpha=0.02
t = 2.387
Lconfint = Slope - t * stndevslope
Uconfint = Slope + t * stndevslope
Print #1, " "
Print #1, "slope = ", slope
Print #1, "intercept = ", incpt
Print #1, "intercept = ", vincpt, "(verify)"
Print #1, "r = ", r
Print #1, "r^2 = ", r * r
Print #1, "Variance = ", Var
Print #1, "Variance of slope = ", varslope
Print #1, "Stnd Deviation of slope = ", stndevslope
Print #1, "98% confidence interval for the slope = ", Lconfint, " to ", Uconfint
Close #1
getnext:
Next N6
Next N5
Next N4
Next N3
Next N2
Next N1
Next N0
Next L7
Next i
Next i5
End Sub
REM: Data from Figge J, Rossing TH, Fencl V. J Lab Clin Med.
REM: 1991; 117:453-467 (Table A).
REM: Data must be entered into a spreadsheet for use in the program.
DATA 01, 7.388, 49.8, 39.3, 1.1, 7.2
DATA 02, 7.383, 45.4, 40.0, 1.0, 7.0
DATA 03, 7.521, 45.4, 26.1, 1.0, 7.0
DATA 04, 7.389, 45.4, 38.1, 1.0, 7.0
DATA 05, 7.217, 45.4, 62.9, 1.0, 7.0
DATA 06, 7.315, 32.2, 23.4, 1.2, 6.6
DATA 07, 7.194, 32.2, 35.0, 1.2, 6.6
DATA 08, 6.979, 32.2, 68.6, 1.2, 6.6
DATA 09, 7.819, 71.3, 28.9, 1.1, 6.8
DATA 10, 7.716, 71.3, 37.7, 1.1, 6.8
DATA 11, 7.504, 71.3, 65.4, 1.1, 6.8
DATA 12, 7.850, 70.2, 26.4, 1.0, 6.8
DATA 13, 7.719, 70.2, 37.9, 1.0, 6.8
DATA 14, 7.513, 70.2, 65.3, 1.0, 6.8
DATA 15, 7.447, 45.9, 30.8, 1.0, 7.1
DATA 16, 7.375, 45.9, 38.4, 1.0, 7.1
DATA 17, 7.094, 45.9, 85.0, 1.0, 7.1
DATA 18, 7.935, 70.2, 22.5, 1.0, 6.8
DATA 19, 7.716, 70.2, 40.2, 1.0, 6.8
DATA 20, 7.423, 70.2, 83.5, 1.0, 6.8
DATA 21, 7.746, 45.8, 22.1, 1.0, 3.5
DATA 22, 7.518, 45.8, 39.9, 1.0, 3.5
DATA 23, 7.218, 45.8, 85.9, 1.0, 3.5
DATA 24, 7.446, 24.2, 21.9, 0.9, 3.4
DATA 25, 7.226, 24.2, 39.8, 0.9, 3.4
DATA 26, 7.018, 24.2, 69.7, 0.9, 3.4
DATA 27, 7.676, 63.7, 40.2, 1.0, 3.6
DATA 28, 7.369, 63.7, 86.7, 1.0, 3.6
DATA 29, 7.711, 75.5, 38.6, 1.0, 6.7
DATA 30, 7.702, 76.4, 38.2, 1.0, 7.3
DATA 31, 7.630, 65.9, 37.9, 1.0, 7.3
DATA 32, 7.572, 60.2, 35.1, 0.7, 7.6
DATA 33, 7.718, 58.9, 41.0, 1.0, 1.7
DATA 34, 7.510, 58.9, 67.9, 1.0, 1.7
DATA 35, 7.399, 58.9, 88.0, 1.0, 1.7
DATA 36, 7.684, 70.4, 38.2, 1.0, 7.0
DATA 37, 7.477, 70.4, 65.4, 1.0, 7.0
DATA 38, 7.390, 70.4, 85.7, 1.0, 7.0
DATA 39, 7.551, 53.5, 38.7, 1.0, 6.2
DATA 40, 7.348, 53.5, 68.4, 1.0, 6.2
DATA 41, 7.240, 53.5, 86.3, 1.0, 6.2
DATA 42, 7.598, 51.2, 41.2, 0.9, 1.9
DATA 43, 7.378, 51.2, 69.8, 0.9, 1.9
DATA 44, 7.307, 51.2, 87.7, 0.9, 1.9
DATA 45, 7.320, 32.5, 22.5, 1.0, 8.0
DATA 46, 7.144, 32.5, 38.9, 1.0, 8.0
DATA 47, 7.006, 32.5, 59.2, 1.0, 8.0
DATA 48, 7.416, 28.5, 22.7, 1.0, 2.9
DATA 49, 7.213, 28.5, 40.3, 1.0, 2.9
DATA 50, 7.068, 28.5, 58.6, 1.0, 2.9
DATA 51, 7.460, 22.8, 23.1, 1.0, 1.6
DATA 52, 7.246, 22.8, 40.2, 1.0, 1.6
DATA 53, 7.083, 22.8, 60.6, 1.0, 1.6
DATA 54, 7.125, 23.7, 22.6, 1.0, 5.7
DATA 55, 6.968, 23.7, 40.0, 1.0, 5.7
DATA 56, 6.849, 23.7, 58.0, 1.0, 5.7
DATA 57, 7.254, 21.4, 23.0, 1.0, 3.5
DATA 58, 7.051, 21.4, 40.7, 1.0, 3.5
DATA 59, 6.924, 21.4, 58.3, 1.0, 3.5
DATA 60, 7.654, 67.5, 39.7, 1.0, 7.2
DATA 61, 7.508, 67.5, 56.9, 1.0, 7.2
DATA 62, 7.347, 67.5, 87.0, 1.0, 7.2
DATA 63, 7.706, 62.5, 40.1, 1.0, 3.8
DATA 64, 7.561, 62.5, 57.5, 1.0, 3.8
DATA 65, 7.386, 62.5, 91.2, 1.0, 3.8
END
Optimized Model Parameters.
Contents of output file: 'model-results':
Checksum1 = 481.218
Checksum2 = 3194.9
Checksum3 = 3210.8
Checksum4 = 65.1
Checksum5 = 342.8
abvs / n = 2.51751200069458E-02
s= -0.221307340304365
ss= 7.00002270873043E-02
HIS14 = 5.1
HIS15 = 6.7
HIS16 = 6.2
LYS1 = 5.800; N1 = 2
LYS2 = 6.150; N2 = 2
LYS3 = 7.510; N3 = 2
LYS4 = 7.685; N4 = 2
LYS5 = 7.860; N5 = 1
LYS6 = 8.035; N6 = 0
LYS7 = 10.30; N7 = 50
slope = 1.00678874215608
intercept = -4.68547320397751E-02
intercept = -4.68547320394102E-02 (verify)
r = 0.991565222939416
r^2 = 0.983201591342893
Variance = 1.09623723925786E-03
Variance of slope = 2.74891760302995E-04
Stnd Deviation of slope = 1.65798600809233E-02
98% confidence interval for the slope = 0.967212616142919 to 1.04636486816925
The sum of squares of the differences between pH (measured) and pH (calculated) is 0.07000.
The optimized parameters are as follows:
The pK(a) s of three histidine residues not determined by NMR spectroscopy:
HIS14 = 5.10
HIS15 = 6.70
HIS16 = 6.20
The low-titrating lysine residues were assigned the following pK(a) values:
LYS1: pK(a) = 5.800; N1 = 2
LYS2: pK(a) = 6.150; N2 = 2
LYS3: pK(a) = 7.510; N3 = 2
LYS4: pK(a) = 7.685; N4 = 2
LYS5: pK(a) = 7.860; N5 = 1
The normally-titrating lysine residues were assigned the following pK(a) value:
LYS7: pK(a) = 10.30
Figge-Fencl Model (Version 3.0) Summary.
The model includes the following features for human serum albumin:
1 Cys residue; pK(a) = 8.5
98 Glu and Asp residues; pK(a) = 3.9
18 Tyr residues; pK(a) = 11.7
24 Arg residues; pK(a) = 12.5
59 Lys residues; 2 with pK(a) = 5.800; 2 with pK(a) = 6.150; 2 with pK(a) = 7.510;
2 with pK(a) = 7.685; 1 with pK(a) = 7.860; and 50 with pK(a) = 10.30
16 His residues; with pK(a)'s of 7.12; 7.22; 7.10; 7.49; 7.01; 7.31; 6.75; 6.36; 4.85; 5.76;
6.17; 6.73; 5.82; 5.10; 6.70; and 6.20 (note that the pK(a)'s of the first five His
residues will each downshift by 0.4 pH units due to the structural rearrangement associated
with the N-B transition).
amino terminus; pK(a) = 8.0
carboxyl terminus; pK(a) = 3.1
Charge on Albumin at pH 7.40 as Calculated per the Model.
At pH 7.40, calculations using the model predict that albumin will carry a net
charge of -18.538 Eq / mol. Therefore, in human plasma at pH 7.40 with [ Albumin ] = 4.4
g/dL, the charge attributed to albumin per the model is -12.3 mEq/L,
calculated as follows:
[ Albx- ] = ( -18.538 Eq/mol ) x ( 10 dL / 1 L ) x ( 1000 mEq / 1 Eq ) x ( 4.4 g/dL )
/ ( 66500 g/mol ) = -12.3 mEq/L;
where 66500 g/mol is the molecular mass of albumin.
Charge on Albumin over the pH Range of 6.9 to 7.9: Linear Regression.
The Figge-Fencl model (version 3.0) predicts that the titration curve of human serum albumin (charge displayed
by albumin, expressed as mEq per g of albumin, versus pH) is approximately
linear over the pH range of biologic interest (6.9 to 7.9) [ see the
albumin titration curve ]. Therefore, a standard
least-squares linear regression algorithm can be used to predict the approximate
charge contributed by albumin in human plasma over this pH interval:
[ Albx- ] = -10 x [ Albumin ] x ( 0.123 x pH - 0.631 );
where [ Albx- ] is in mEq/L and [ Albumin ] is in g/dL. Therefore, at pH 7.40, the
charge contributed by 4.40 g/dL of albumin is approximately -12.3 mEq/L.
Molar Buffer Capacity of Albumin as Calculated per the Model.
The molar buffer capacity of albumin is calculated by taking the additive inverse of the first derivative
of [ Albx- ] with respect to pH:
- d [ Albx- ] / d pH
At pH = 7.40, d [ Albx- ] / d pH = -8.2; hence, the molar buffer capacity of albumin is
8.2 Eq/mol/pH. This value is similar to the value of 7.6 Eq/mol/pH calculated by Fogh-Andersen (1993)
Thus, for [ Albumin ] of 4.4 g/dL, the buffer value of albumin at pH 7.40 is:
( 4.4 g/dL ) x ( 8.2 Eq/mol/pH ) x ( 1 mol / 66500 g ) x ( 1000 mEq / 1 Eq ) x ( 10 dL / 1 L ) =
5.4 mEq/L/pH
The average value of the molar buffer capacity of albumin over the pH range of 6.9 to 7.9 can be calculated
using the slope of the least-squares regression line over that interval:
( 0.123 mEq/g/pH ) x ( 66500 g/mol ) x ( 1 Eq / 1000 mEq ) = 8.2 Eq/mol/pH
Hence, the average molar buffer capacity over this pH interval is equal to the specific value at the
midpoint (pH = 7.40).
The approximate linearity of the albumin titration curve over the pH range of biologic interest (6.9 - 7.9) is
critical for the successful derivation of the classic van Slyke equation for plasma.
As explained in the online supplement to Fencl, Jabor, Kazda, and Figge (2000)
[ Online Supplement ],
the molar buffer capacity of albumin is used in the formulation of the van Slyke
equation for human plasma. More information about the van Slyke equation can be found on
O. Siggaard-Andersen's web site at
https://www.siggaard-andersen.dk/.
Formal Mathematical Representation of the Figge-Fencl Quantitative Physicochemical Model
of Human Acid-Base Physiology in Blood Plasma (Version 3.0).
Copyright 2003 - 2026 James J. Figge.
Published 8 October, 2012; update published 28 April, 2013; update published 27 October, 2013.
SID + 1000 x ( (aH+) - Kw / (aH+) - Kc1 x PCO2 / (aH+)
- Kc1 x Kc2 x PCO2 / (aH+)2 )
- [ Pitot ] x Z
+ { -1 / ( 1 + 10 ^ - ( pH - 8.5 ) )
- 98 / ( 1 + 10 ^ - ( pH - 3.9 ) )
- 18 / ( 1 + 10 ^ - ( pH - 11.7 ) )
+ 24 / ( 1 + 10 ^ + ( pH - 12.5 ) )
+ 2 / ( 1 + 10 ^ + ( pH - 5.80 ) )
+ 2 / ( 1 + 10 ^ + ( pH - 6.15 ) )
+ 2 / ( 1 + 10 ^ + ( pH - 7.51 ) )
+ 2 / ( 1 + 10 ^ + ( pH - 7.685 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 7.86 ) )
+ 50 / ( 1 + 10 ^ + ( pH - 10.3 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 7.12 + NB ) )
+ 1 / ( 1 + 10 ^ + ( pH - 7.22 + NB ) )
+ 1 / ( 1 + 10 ^ + ( pH - 7.10 + NB ) )
+ 1 / ( 1 + 10 ^ + ( pH - 7.49 + NB ) )
+ 1 / ( 1 + 10 ^ + ( pH - 7.01 + NB ) )
+ 1 / ( 1 + 10 ^ + ( pH - 7.31 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 6.75 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 6.36 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 4.85 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 5.76 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 6.17 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 6.73 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 5.82 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 5.10 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 6.70 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 6.20 ) )
+ 1 / ( 1 + 10 ^ + ( pH - 8.0 ) )
- 1 / ( 1 + 10 ^ - ( pH - 3.1 ) ) } x 1000 x 10 x [ Albumin ] / 66500 = 0.
Where:
(aH+) = 10-pH ; (aH+) is the hydrogen ion activity, also
used as an approximation of hydrogen ion concentration, [H+];
Z = ( K1 x (aH+)2 + 2 x K1 x K2 x (aH+) + 3 x K1 x K2 x K3 ) /
( (aH+)3 + K1 x (aH+)2 + K1 x K2 x (aH+) + K1 x K2 x K3 );
NB = 0.4 x ( 1 - 1 / ( 1 + 10 ^ ( pH - 6.9 ) ) );
Strong Ion Difference, SID, is given in mEq / L;
PCO2 is given in Torr;
Total concentration of inorganic phosphorus-containing species,
[ Pitot ], is given in mmol / L;
[ Albumin ] is given in g / dL;
Kw = 4.4E-14 ( Eq/L )2;
Kc1 = 2.44E-11 ( Eq/L )2 / Torr;
Kc2 = 1.1E-10 ( Eq/L );
K1 = 1.22E-2 ( mol/L );
K2 = 2.19E-7 ( mol/L );
K3 = 1.66E-12 ( mol/L );
66500 g/mol is the molecular mass of albumin.
The above expression defines a function, fpH, which can be used to calculate the
pH of plasma for any valid set of values for SID, PCO2, [ Pitot ],
and [ Albumin ]:
pH = fpH { SID, PCO2, [ Pitot ], [ Albumin ] }
The function is too complex to be solved by hand and must be solved via an iterative
approach on a computer.