Amino Acid Side-chain
Hydrophobicity
Hue Sun Chan, University of Toronto, Toronto, Canada
Hydrophobicity is the unfavourable energetics of dissolving nonpolar compounds in
water. The hydrophobicities of the 20 amino acid side-chains are currently described by
hydrophobicity scales derived primarily from solubility studies; these scales have provided
semiquantitative rationalizations of some properties of native (folded) proteins.
Introduction
Hydrophobicity is a term used to describe the unfavourable
energetics of dissolving nonpolar (e.g. carbon)
compounds in water: while nonpolar compounds readily
dissolve in many nonpolar solvents, they have very low
solubilities in water. Another defining characteristic of
hydrophobic compounds is their tendency to self-associate
in aqueous solutions. A generally accepted explanation of
this phenomenon is that it results largely from the more
favourable interactions among the water molecules
achieved when the nonpolar groups cluster together,
rather than from a direct van der Waals attraction between
nonpolar groups.
What is the physical origin of hydrophobicity? There are
extensive hydrogen-bonding interactions among water
molecules (H2O) in liquid water. Hydrophobic groups,
being nonpolar, cannot form hydrogen bonds with the
water molecules in their vicinity. However, it would be
energetically costly if these water molecules are not
hydrogen bonded; therefore, in the presence of a nonpolar
solute, water molecules are driven to either adopt more
specific orientations to avoid losing hydrogen bonds, or to
minimize the loss, depending on which alternative is
allowed by the geometry of the hydrophobic surface
(Figure1). The presence of special water configurations near
a hydrophobic surface, with average properties distinguishable
from that of bulk water, is often referred to as
‘hydrophobic hydration’. The statistical mechanical consequences
are either a restriction of water configurational
freedom (entropy) or a loss of favourable (enthalpic)
hydrogen bonds, or both. These effects translate into an
unfavourable free energy change associated with exposing
hydrophobic groups to water (Figure 1).
Liquid state configurations are constantly fluctuating,
not static. Figure 1 illustrates only the dominant configurations;
it is possible to have other water configurations near
hydrophobic surfaces. However, the dominant configurations
in Figure 1 show how unfavourable free energy
changes are caused by hydrophobic surfaces’ disruption of
water structure relative to that in the bulk. Consistent with
this molecular picture, it has been empirically observed
that the magnitude of hydrophobic effect, as measured by
transfer free energies (see below), is approximately
proportional to the water-accessible surface area (Lee
and Richards, 1971) of the hydrophobic groups (Tanford,
1980). In this perspective, the association of hydrophobic
groups in water is viewed as driven by the need to reduce
energetically unfavourable exposure of nonpolar surface
to water. Typically, transferring hydrophobic groups from
water to a nonpolar environment is believed to be favoured
by approximately 105–142 J mol2 1
(25–34 cal mol2 1
) per
square angstrom of water-accessible area of the hydrophobic
group (reviewed by Chan and Dill, 1997). (In the
biophysics literature, water-accessible surface area is often
called ‘solvent-accessible surface area’ (SASA) without
specifying explicitly that water is the solvent in question.)
Experimentally Derived Scales from
Free Energies of Transfer
It is a common practice to use experimentally determined
free energies of transfer of various solutes, from nonpolar
solvents to water, to quantify the extent of hydrophobicity,
i.e. the degree to which each solute ‘dislikes’ water, relative
to the nonpolar phase (Figure 2). In these measurements, a
solute is partitioned between two different solvents
according to its solubility in each. The experimental
observable is the equilibrium concentration of the solute
in each phase. Transfer free energies are determined from
the measured concentrations, and contact energies are
deduced using thermodynamic arguments and statistical
mechanical models (reviewed by Chan and Dill, 1997).
Because the process of folding a protein involves
removing a large fraction of amino acid surface from
water to the protein interior, it is on some level analogous
to transferring a model compound from water to a
nonpolar phase. Consequently, many researchers have
obtained experimental transfer data for amino acids, their
derivatives and other model compounds with the goal of
using such data to understand the origins of protein
stability. While the analogy to water phase is direct,
Article Contents
Secondary article
. Introduction
. Experimentally Derived Scales from Free Energies of
Transfer
. Dependence of Experimental Scale on Phase, Solvent
. Theoretical Scales Derived from Protein Structures
. Comparison Between Prediction and Experiment
1ENCYCLOPEDIA OF LIFE SCIENCES © 2002, John Wiley & Sons, Ltd. www.els.net
discerning which nonpolar phase best models a protein
interior has been less obvious. As a result, a variety of
different nonpolar phases have been employed. These
include octanol, linear alkanes, cyclohexane, bilayer,
grafted alkyl phase on chromatographic columns, among
others.
Hydrophobicity scales have been determined from free
energies of transfer of amino acids or their chemical
derivatives from water to various nonpolar phases (for
example, an organic solvent, see Figure 2). In analysing
transfer data, it is often assumed that contributions to the
transfer free energy from different parts (e.g. a methyl or
amide group) of a molecule are additive. When group
additivity is assumed, contributions to the free energy of
transfer of individual amino acids can be calculated from
differences in transfer free energies of various amino acid
derivatives. By the same token, some scales give hydrophobicities
of individual amino acids in terms of whether
they are more hydrophobic or less hydrophobic than
glycine, the smallest amino acid that does not have a sidechain
(Table 1). The assumption of group additivity is quite
reasonable, and is often useful. None the less, it should be
noted that, ultimately, group additivity and treatments
based solely on water-accessible surface area are only
approximations.Calculationsattheatomiclevel(Leeetal.,
1984) show that the underlying molecular basis of
hydrophobicity is more intricate than these simplified
descriptions.
Dependence of Experimental Scale on
Phase, Solvent
Given the central importance of understanding what
determines protein stability and structure, numerous
amino acid hydrophobicity scales have been obtained by
many research groups during the last few decades.
Representative scales are given in Table 1. Information
regarding other scales and recent reviews is provided in
Table 2. In general, there has been a lack of quantitative
agreement between hydrophobicity scales. Table 1 demonstrates
that different nonpolar phases and different
techniques give rise to different amino acid transfer free
energies. Figure 3a provides a visual comparison of the level
of quantitative agreement between scales.
Figure1 Molecularoriginsofhydrophobicity.Typicalhydrogenbonding(dashedlines)patternamongwatermoleculesH2O:(a)in thebulk;(b)arounda
smallnonpolarsolute (shadedcircle);and(c) nearanextendednonpolarsurface.Thehydrogenbondinggeometryin (b)isdistorted relativetothat in(a)to
maintain an interaction strength among water molecules comparable to that in the bulk. Water molecules around a small solute with a convex nonpolar
surface are oriented to avoid directing their hydrogen-bonding groups (donor or acceptor) towards the solute. This arrangement is not possible near a flat
extended nonpolar surface. In this case thereare ‘dangling’hydrogen bonds, i.e. potentially hydrogen-bonding groups (dottedlines) oriented towards the
nonpolar surface (Lee et al., 1984).
Amino Acid Side-chain Hydrophobicity
2
The charged and polar (hydrogen-bonding) side-chains
are particularly sensitive to the nonpolar phase, and their
transfer free energies show large deviation from the
expected proportionality to water-accessible surface area.
This may be caused by a number of factors. For instance,
octanol, one of the commonly used nonpolar phase
solvents, is somewhat polar and contains a significant
amount of dissolved water, whereas other nonpolar phase
solvents, such as liquid alkanes, have very little dissolved
water. On the other hand, the hydrophobicities of the
hydrophobic/nonpolar amino acid side-chains (aliphatic
nonhydrogen bonding, aromatic nonhydrogen bonding
and sulfur-containing) show less variation among different
scales (Karplus, 1997). The temperature dependence of
transfer free energies can also vary significantly. In some
experiments involving partially aligned alkyl chains, such
as reverse-phase liquid chromatography stationary phases
(Figure 2c), nonpolar solutes are driven mainly by enthalpy
into the partially aligned alkyl phase, instead of by entropy
as in transfer between water and bulk nonpolar phases.
This observation is sometimes called ‘nonclassical’ hydrophobic
effect or ‘bilayer effect’ (discussed in DeVido et al.
(1998) and White and Wimley (1999); Table 2).
Given the variability between different hydrophobicity
scales, it is perhaps not surprising that the rank ordering of
amino acid hydrophobicities shows marked variation from
scale to scale. For example, in the 16 scales tabulated by
Wilce et al. (1995), phenylalanine is ranked first (most
hydrophobic) by four of the scales, yet one scale ranks it as
16th, i.e. close to being the least hydrophobic (20th). In
Table 1, tryptophan is ranked first by two scales (b and c),
whereas it is the third most hydrophobic according to scale
(a). These observations highlight the limitations of amino
acid hydrophobicity scales in quantitative applications.
Unfortunately, no one set of 20 numbers (i.e. a single
generic hydrophobicity scale) exists that is capable of
accurately predicting protein stability and/or ligandbinding
energetics.
Theoretical Scales Derived from Protein
Structures
As discussed above, most hydrophobicity scales use a
nonpolar phase to model a generic protein core. As such,
they cannot account for the specific interactions between
amino acid side-chains in protein interiors.
In principle,these interactionscould bestudied ata more
fundamental level by using potentials for each atom.
However, for many applications, an amino acid-based
approach is still preferred, because a calculation involving
the pairwise interactions between the thousands of atoms
in a given protein is often not tractable with currently
available computational power. A logical next step,
therefore, is to determine a set of 210 pairwise amino acid
contact energies.
To date, no systematic, direct experimental measurements
of all pairwise amino acid contact energies exists.
(a)
(b)
(c)
(d)
Figure 2 Experimental and statistical procedures for estimating amino
acid interactionparameters.(a) Formationof a contactwhena biomolecule
undergoes conformational changes, as in protein folding. The aqueous
solvent is not depicted explicitly. (b) Modelling contact formation by
transferring a model compound (small solute) from an aqueous phase (left)
to a nonpolar phase (right). Hydrophobic hydration is also studied by
transferring small solutes from a gaseous phase (middle) to water. (c)
Modelling contact formation by studying the partitioning of solutes into
aligned nonpolar phases such as bilayers and in reversed-phase liquid
chromatography experiments. (d) Some interaction parameters are
deduced from the statistics of contacts among different amino acid types in
the database of protein native structures.
Amino Acid Side-chain Hydrophobicity
3
Instead, these energies have been obtained by knowledgebased
or statistical methods from databases of protein
native structures. (Hence they are called statistical
potentials.) There are a number of slightly different
approaches, but they all share the same basic assumption
that the statistical distribution of contacts among native
structures of proteins can be related to the underlying
contact energies by some very simple mathematical
relations (Figure 2d). Tanaka and Scheraga first advanced
this idea in 1976. A comprehensive analysis was provided
by Miyazawa and Jernigan in 1985 (Miyazawa and
Jernigan, 1996), and subsequently developed by many
researchers.
The formulation of Miyazawa and Jernigan (1996) is a
good illustration of the statistical potential method. They
define the energy for a contact between amino acid residue
types i and j by eij 5 2 RT ln [(nnijnn00)/( nni0nnj0)], where nns are
average numbers of contacts observed in a given database
of native structures. The subscript 0 denotes solvents, 00
and i0 represent a solvent–solvent contact and a contact
between residue type i and solvent (when i is exposed to
solvent), respectively. By this simple formula, more
favourable (lower) contact energies eij are assigned to
pairs of amino acid side-chains that occur more frequently
together in spatial contact (i.e. larger nnij) in the given
protein native structure database.
This and similar approaches assume that the contacts in
the protein native structure database follow a Boltzmann
distribution at temperature T. In reality, however, the set
of native structures is not a Boltzmann ensemble. Native
structures of different proteins are not in thermal
equilibrium because they cannot interconvert into one
another. Nevertheless, in test studies using simple model
systems, these simple statistical procedures have been
shown to be reasonably accurate in extracting contact
energies, provided that certain special conditions are
satisfied. It should also be pointed out, however, that these
procedures have been shown to be not generally valid. A
rigorous connection has yet to be established between
statistical potentials and the true physical interactions.
Table 1 Free energies of transfer of the amino acids from water to nonpolar environments,
relative to that of glycine
a
These hydrophobicity scales were obtained by using different chemical derivatives of the amino acids and the
following different nonpolar phases: (a) alkyl chains in a reversed-phase liquid chromatography stationary
phase (DeVido et al., 1998), (b) bulk octanol (Fauchère and Pliška, 1983), and (c) large unilamellar vesicle
membranes (Wimley and White, 1996; with ionized side-chains E, D, R, K and H).
Free energy transfera (kJ mol–1)
Amino acid Code (a) (b) (c)
Phenylalanine F –12.3 –10.1 –4.79
Isoleucine I –11.4 –10.2 –1.34
Tryptophan W –11.3 –12.8 –7.81
Leucine L –8.95 –9.64 –2.39
Tyrosine Y –8.70 –5.44 –3.99
Methionine M –8.23 –6.97 –1.01
Valine V –7.80 –6.91 +0.25
Proline P –7.15 –4.08 +1.85
Cysteine C –4.05 –8.73 –1.05
Glutamate E –3.02 +3.63 +8.44
Alanine A –2.67 –1.76 +0.67
Threonine T –2.50 –1.47 +0.55
Glutamine Q –1.95 +1.25 +2.39
Aspartate D –1.03 +4.36 +5.12
Glycine G 0.00 0.00 0.00
Serine S +0.42 +0.23 +0.50
Asparagine N +1.25 +3.40 +1.72
Arginine R +1.75 +5.72 +3.36
Lysine K +3.00 +5.61 +4.12
Histidine H +4.22 –0.74 +3.99
Amino Acid Side-chain Hydrophobicity
4
Comparison Between Prediction and
Experiment
What can be learned about protein structure and stability
from hydrophobicity scales and pairwise amino acid
statistical potentials? At a qualitative or semiquantitative
level, predictions by some scales are consistent with protein
experiments. For example, amino acid residues that are
deemed more hydrophobic in transfer experiments are
more likely to be buried in native structures of globular
proteins. In membrane proteins, contiguous stretches of
hydrophobic residues are useful for identifying membranespanning
segments. More quantitatively, Figure 3b shows
that knowledge-based statistical potentials correlate well
with hydrophobicities of individual amino acids determined
by nonpolar/water transfers. The statistical potentials
are derived from experimentally determined protein
native structures. Hence, the correlation in Figure 3b
indicates that some scales derived from model compound
transfer experiments are predictive, in the sense that they
are able to provide a reasonable account of the structural
organization of native proteins (see also Rose et al. (1985)
and Eisenberg and McLachlan (1986); Table 2).
Amino acid hydrophobicity scales have been used to
rationalize changes in protein stability caused by changes
in amino acid sequence (mutations); however, the results
are not straightforward to interpret. Figure 3c shows that
there are two sources of uncertainties in relating experimental
mutagenesis data to hydrophobicity scales: (1)
variation between different hydrophobicity scales, as noted
above; and (2) variation in folding free energy changes
caused by mutations between the same two amino acid
types, which turn out to depend on the protein and the
particular location of the mutation site. This observation
implies that the protein environment of a given amino acid
residue can strongly affect its interactions. Twentyparameter
hydrophobicity scales cannot accurately predict
mutational effects on native stability because these scales
effectively assume a single uniform generic protein core
environment.
–8
–17
–17 101
Freeenergyoftransfer(kJmol–1)
(anRPLCscale)
Free energy of transfer (kJ mol–1) (other scales)
7
(a)
–11 74
10
–14 –8 –5 –2
–14
–11
–5
–2
4
1
–15
–30
–8 0–2
Pairwisesumoffreeenergiesoftransfer
(FauchereandPliska,1983)
Miyazawa and Jernigan (1996) interaction parameters
10
(b)
–6 –1
15
–7 –5 –4 –3
–25
–20
–10
–5
5
0
12
0
∆∆Gfold
(kJmol–1)
27
(c)
33
3
9
15
18
24
21
6
30
I — V I — A I — G L — A L — G V — A V — G
´
ˆ
Figure 3 Hydrophobicity scales obtained from different techniques and
their applications to protein folding. (a) Correlation between the reversedphase
liquid chromatography (RPLC) scale in Table 1(a) and one of the
Wimley, Creamer and White (1996) scales (red dots), the scale of Fauche` re
and Plis˘ ka (1983) (blue dots, Table 1 (b)), and that of Wimley and White
(1996) (green dots, Table 1 (c)). Least squarefits are given by the upper and
lower solid lines and the dashed line, with correlation coefficient r 5 0.96,
0.87 and 0.72, respectively; see DeVido et al. (1998) in Table 2 for details.
(b) A set of 210 interaction parameters between pairs of amino acids
determined statistically from a database of protein native structures
(Miyazawa and Jernigan, 1996) are plotted against pairwise sums of
hydrophobicities (Fauche` re and Plis˘ ka, 1983) of the corresponding amino
acids; solid line is the least square fit, r 5 0.90. (c) DDGfold is the folding freeenergy
change caused by the type of single-site mutation given below the
horizontal axis. A larger DDGfold means that the mutation results in a less
stablenativestructure.Atotal of48differentexperimentalvaluesofDDGfold
are plotted (open circles). The same mutation can produce very different
changes in folding free energy in different proteins or at different sites of the
same protein. The ranges of corresponding free-energy changes predicted
by small model compound results and the analysis of Lee are indicated by
the dashed boxes. Part (c) of this figure is adapted from Lee (1993); more
details are given in this reference.
3———————————————————————
Amino Acid Side-chain Hydrophobicity
5
Statistical potentials have been used in fold identification
for predicting protein structure from sequence (Sippl,
1995). Given an amino acid sequence, a fold identification
technique attempts to recognize the native structure
among alternate folds (decoys). Typically, a total energy
or score is computed as a sum of pairwise contact energies
for each of the conformations to be considered. The
technique is successful if the native conformation has an
energy lower (more favourable) than all the decoys. When
these procedures are tested using protein sequences with
known native structures, they often fail when the decoy set
used to challenge the identification technique contains a
wide range of compact conformations. These failures
suggest that some essential aspects of protein interactions
are missing in current pairwise contact energies.
In summary, while low-resolution amino acid-based
contact energies and hydrophobicity scales have provided
insight into protein energetics, their utility in quantitative
predictions has proved limited. Advances in both experimental
and theoretical treatments beyond these empirical
approaches are needed to provide more accurate accounts
of the interactions among amino acid side-chains, and
hence a more detailed understanding of protein stability
and structure.
References
Chan HS and Dill KA (1997) Solvation: how to obtain microscopic
energies from partitioning and solvation experiments. Annual Review
of Biophysics and Biomolecular Structure 26: 425–459.
DeVido DR, Dorsey JG, Chan HS and Dill KA (1998) Oil/water
partitioning has a different thermodynamic signature when the oil
solvent chains are aligned than when they are amorphous. Journal of
Physical Chemistry B 102: 7272–7279.
Eisenberg D and McLachlan AD (1986) Solvation energy in protein
folding and binding. Nature 319: 199–203.
Eriksson AE, Baase WA, Zhang X-J et al. (1992) Response of a protein
structure to cavity-creating mutations and its relation to the
hydrophobic effect. Science 255: 178–183.
Fauche` re J-L and Plis˘ ka V (1983) Hydrophobic parameters P of aminoacid
side chains from the partitioning of N-acetyl-amino-acid amides.
European Journal of Medicinal Chemistry – Chimie The´rapeutique 18:
369–375.
Guo D, Mant CT, Taneja AK et al. (1986) Prediction of peptide
retention times in reversed-phase high-performance liquid chromatography.
I. Determination of retention coefficients of amino acid
residues of model synthetic peptides. Journal of Chromatography 359:
499–517.
Karplus PA (1997) Hydrophobicity regained. Protein Science 6: 1302–
1307.
Kim A and Szoka FC Jr (1992) Amino acid side-chain contributions to
free energy of transfer of tripeptides from water to octanol.
Pharmaceutical Research 9: 504–514.
Lee B (1993) Estimation of the maximum change in stability of globular
proteins upon mutation of a hydrophobic residue to another of smaller
size. Protein Science 2: 733–738.
Lee B and Richards FM (1971) The interpretation of protein structures:
estimation of static accessibility. Journal of Molecular Biology 55: 379–
400.
Lee CY, McCammon JA and Rossky PJ (1984) The structure of liquid
water at an extended hydrophobic surface. Journal of Chemical
Physics 80: 4448–4455.
Meek JL and Rosetti ZL (1981) Factors affecting retention and
resolution of peptides in high-performance liquid chromatography.
Journal of Chromatography 211: 15–28.
Table 2 References for representative hydrophobicity scales obtained by different experimental methods
Experimental method Reference
Transfer experiments Nozaki and Tanford (1971) reviewed in Tanford (1980)
(Table 13-1, p. 140)
Fauche` re and Plis˘ ka (1983) (see Table 1)
Kim and Szoka (1992)
Wimley, Creamer and White (1996)
Comparing protein structure data with transfer experiments Rose et al. (1985)
Eisenberg and McLachlan (1986)
Mutation studies Yutani et al. (1987)
Eriksson et al. (1992)
Bilayer studies Wimley and White (1996) (see Table 1)
Thorgeirsson et al. (1996)
White and Wimley (1999)
Chromatography Meek and Rosetti (1981)
Guo et al. (1986)
Wilce et al. (1995)
DeVido et al. (1998) (see Table 1)
More comprehensive lists of hydrophobicity scales can be found in Wilce et al., 1995 (16 scales are compared) and DeVido et al., 1998 (more
than 40 scales are considered and cited in Tables 3 and 4 of this reference).
Amino Acid Side-chain Hydrophobicity
6
Miyazawa S and Jernigan RL (1996) Residue–residue potentials with a
favorable contact pair term and an unfavorable high packing density
term, for simulation and threading. Journal of Molecular Biology 256:
623–644.
Rose GD, Geselowitz AR, Lesser GJ et al. (1985) Hydrophobicity of
amino acid residues in globular proteins. Science 229: 834–838.
Sippl MJ (1995) Knowledge-based potentials for proteins. Current
Opinion in Structural Biology 5: 229–235.
Tanford C (1980) The Hydrophobic Effect: Formation of Micelles and
Biological Membranes, 2nd edn. New York: Wiley.
Thorgeirsson TE, Russell CJ, King DS and Shin YK (1996) Direct
determination of the membrane affinities of individual amino acids.
Biochemistry 35: 1803–1809.
White SH and Wimley WC (1999) Membrane protein folding and
stability: physical principles. Annual Review of Biophysics and
Biomolecular Structure 28: 319–365.
Wilce MCJ, Aguilar M-I and Hearn MTW (1995) Physiochemical basis
of amino acid hydrophobicity scales: evaluation of four new scales of
amino acid hydrophobicity coefficients derived from RP-HPLC of
peptides. Analytical Chemistry 67: 1210–1219.
Wimley WC, Creamer TP and White SH (1996) Solvation energies of
amino acid side chains and backbone in a family of host–guest
pentapeptides. Biochemistry 35: 5109–5124.
Wimley WC and White SH (1996) Experimentally determined hydrophobicity
scale for proteins at membrane interfaces. Nature Structural
Biology 3: 842–848.
Yutani K, Ogasahara K, Tsujita T and Sugino Y (1987) Dependence of
conformational stability on hydrophobicity of the amino acid residue
in a series of variant proteins substituted at a unique position of
tryptophan synthase a subunit. Proceedings of the National Academy
of Sciences of the USA 84: 4441–4444.
Further Reading
Cheng Y-K and Rossky PJ (1998) Surface topography dependence of
biomolecular hydrophobic hydration. Nature 392: 696–699.
Dill KA (1990) Dominant forces in protein folding. Biochemistry 29:
7133–7155.
Hummer G, Garde S, Garcia AE, Paulaitis ME and Pratt LR (1998)
Hydrophobic effects on a molecular scale. Journal of Physical
Chemistry B 102: 10469–10482.
Kyte J and Doolittle RF (1982) A simple method for displaying the
hydropathic character of a protein. Journal of Molecular Biology 157:
105–132.
Liu L-P and Deber CM (1998) Uncoupling hydrophobicity and helicity
in transmembrane segments: a-helical propensities of the amino acids
in non-polar environments. Journal of Biological Chemistry 273:
23645–23648.
Park BH, Huang ES and Levitt M (1997) Factors affecting the ability of
energy functions to discriminate correct from incorrect folds. Journal
of Molecular Biology 266: 831–846.
Privalov PL and Gill SJ (1988) Stability of protein structure and
hydrophobic interaction. Advances in Protein Chemistry 39: 191–234.
Robertson AD and Murphy KP (1997) Protein structure and the
energetics of protein stability. Chemical Reviews 97: 1251–1267.
Scheraga HA (1998) Theory of hydrophobic interactions. Journal of
Biomolecular Structure and Dynamics 16: 447–460.
Skolnick J, Jaroszewski L, Kolinski A and Godzik A (1997) Derivation
and testing of pair potentials for protein folding. When is the
quasichemical approximation correct? Protein Science 6: 676–688.
Amino Acid Side-chain Hydrophobicity
7