Computational protein design promises to revolutionize protein engineering.

Abstract

Natural evolution has produced an astounding array of proteins that perform the physical and chemical functions required for life on Earth. Although proteins can be reengineered to provide altered or novel functions, the utility of this approach is limited by the difficulty of identifying protein sequences that display the desired properties. Recently, advances in the field of computational protein design (CPD) have shown that molecular simulation can help to predict sequences with new and improved functions. In the past few years, CPD has been used to design protein variants with optimized specificity of binding to DNA, small molecules, peptides, and other proteins. Initial successes in enzyme design highlight CPD's unique ability to design function de novo. The use of CPD for the engineering of potential therapeutic agents has demonstrated its strength in real-life applications.

Full text

Page 1

Vol. 42 ı No. 1 ı 2007 www.biotechniques.com ı BioTechniques ı 31
Tech Insight
Computational protein design promises to revolutionize protein engineering
Oscar Alvizo1, Benjamin D. Allen2, and Stephen L. Mayo3
1Biochemistry and Molecular Biophysics Option, 2Division of Chemistry and Chemical Engineering, and 3Howard Hughes Medical Institute, Division of
Biology and Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, CA, USA
Introduction
Proteins are linear heteropolymers, synthesized from 20
different amino acids. They participate in nearly all of the
structural, catalytic, sensory, and regulatory functions that
are associated with living systems; these myriad roles are
made possible by their ability to self-assemble into well-
defined structures specified by their sequences. Molecu-
lar evolution has shown that varying a protein sequence
through mutation and recombination can generate new
structures and functions. Drexler first noted that exist-
ing biological systems for protein fabrication could be
harnessed to produce nanoscale molecular machines with
designed functions (1). However, the potential of this ap-
proach can only be fully realized with reliable methods to
predict sequences that perform the desired functions. Over
the last decade, computational protein design (CPD) has
clearly shown its potential as a solution to this problem.
CPD was first conceived as the inverse of the protein-
folding problem, since its goal is to generate amino acid
sequences that adopt a specific three-dimensional fold (2).
CPD utilizes the main-chain coordinates of a known protein
structure as a fixed scaffold. Various amino acid types are
modeled at each designed position, and potential muta-
tions are suggested based on their interactions with the
scaffold and with each other (Figure 1A).
Although the main chain is held fixed during a CPD cal-
culation, various conformations of each amino acid type
at each position are sampled to find sequences expected
to stabilize the fold and satisfy any additional functional
requirements. The distribution of energetically accessible
conformations available to each amino acid side-chain
is approximated using a set of discrete, low-energy con-
formations called rotamers (3,4) (Figure 1, B–D). At the
beginning of a typical CPD calculation, rotamers of the
user-specified amino acid types are assigned to residue
positions from a predefined library. The problem is thus to
find a choice of rotamer at each position such that the fold
is stabilized and the desired function is achieved.
Energy Functions
Amino acid sequences and conformations are scored us-
ing a set of energy functions designed to reproduce the
features of stable proteins. Although the specific energy
functions used and their parameters vary between differ-
ent CPD implementations, most implementations include
a function that prevents atomic overlap and favors van
der Waals interactions (5) and a function that benefits the
formation of hydrogen bonds (6,7). Although interactions
between a protein and its aqueous environment are crucial
for stability, it would be prohibitively expensive to model
water molecules explicitly in a CPD calculation. Therefore,
solvation potentials are used to reward the burial of hy-
drophobic groups and penalize the burial of polar groups;
energies are computed using surface area (8,9) or occlud-
ed volume models (10,11). Electrostatic interactions may
be modeled using Coulomb’s law with a constant dielectric
(6,12), a statistical pair potential (13), or more sophisticat-
ed methods (14). These energy functions were designed to
simulate different conformations of a single sequence and
can give spurious results when used to choose between
different sequences. Therefore, the scoring functions are
typically supplemented with heuristic, statistical, and
negative design terms to compensate for the limitations of
the inverse folding model. These terms include heuristic
estimates of side-chain entropy (15), penalties for non-
polar exposure (5), statistical rotamer probabilities (13),
and composition-based unfolded state energies (13,15).
Proteins have been successfully designed with multiple
Natural evolution has produced an astounding array of proteins that perform the physical and chemical functions
required for life on Earth. Although proteins can be reengineered to provide altered or novel functions, the utility
of this approach is limited by the difficulty of identifying protein sequences that display the desired properties.
Recently, advances in the field of computational protein design (CPD) have shown that molecular simulation can
help to predict sequences with new and improved functions. In the past few years, CPD has been used to design
protein variants with optimized specificity of binding to DNA, small molecules, peptides, and other proteins. Initial
successes in enzyme design highlight CPD’s unique ability to design function de novo. The use of CPD for the
engineering of potential therapeutic agents has demonstrated its strength in real-life applications.

Page 2

Vol. 42 ı No. 1 ı 2007 www.biotechniques.com ı BioTechniques ı 33
Tech Insight
combinations of these functions, but no consensus has yet
been reached on the ideal set of functions or the proper
weight for each term.
Sequence Optimization
The energy functions in CPD are used to compute the
interaction energies between each pair of rotamers at
different designed positions (pair energies), as well as the
interaction energy between each rotamer and the fixed
scaffold (singles energies). Given a rotamer at each posi-
tion, the total energy can be computed by summing the
singles for each rotamer and pairs for each pair of rotam-
ers. Typically, all singles and pair energies are precom-
puted, yielding a combinatorial optimization problem for
which the minimum energy solution corresponds to the
amino acid sequence and conformation expected to best
stabilize the fold. Several optimization algorithms have
been used to solve the design problem. Monte Carlo with
simulated annealing (16–18), genetic algorithms (18,19),
and fast and accurate side-chain topology and energy
refinement (FASTER) (20,21) are stochastic optimization
routines that sequentially improve an existing solution
by making perturbations and accepting or rejecting them
based on their energies. These algorithms are able to gen-
erate reasonable solutions quickly and will always give an
answer, regardless of the difficulty of the design problem.
Their running times can be extended indefinitely in an
attempt to improve the quality of the prediction. However,
they can never guarantee that the
optimal sequence has been found.
Alternatively, strategies based on the
dead-end elimination (DEE) theorem
remove singles and pairs from the
calculation that can be mathemati-
cally proven not to exist in the opti-
mal solution (22,23). Although DEE
can find optimal solutions to smaller
problems, it does not scale well, and
one must resort to the inexact algo-
rithms described above when it fails
to yield a definitive answer.
Design of Function
Although a designed protein se-
quence must adopt the desired fold,
most design goals require that some
useful function be achieved as well.
To this end, the scoring functions de-
scribed above are sometimes extend-
ed with additional function-specific
constraints (24–27). For example, the
geometric and chemical requirements
for ligand or transition-state binding
in designed sensors or enzymes may be too difficult to
model using standard potential functions. Thus, a theoreti-
cal model of the appropriate ligand contacts may be gener-
ated using quantum mechanics or chemical intuition, and
the chemical and geometric features of this model can be
enforced explicitly during the design. Typically, this type of
model will specify the required contacts between the pro-
tein and the ligand, the types of protein groups allowed to
make each contact, and a range of acceptable geometries
for each interaction.
Applications of CPD
CPD was initially used as a tool to improve our understand-
ing of the intrinsic properties of proteins and focused
primarily on maintaining or enhancing protein stability.
However, in recent years, the use of larger scaffolds and
a focus on function have moved the CPD field into a new
chapter. It is now more often used to produce proteins with
novel or modified functions.
Ultimately, the CPD field would like to be able to rationally
engineer any function into a given scaffold. A scaffold is
usually obtained from the crystal structure of a naturally
occurring protein; however, it is possible to computation-
ally design one from the ground up (13). Recent work has
focused on a handful of common biological functions, such
as the interaction of proteins with small molecules, nucleic
acids, and other proteins. In addition, major efforts toward
the de novo design of catalytic activity are underway. CPD
Figure 1. The inverse folding design model. (A) A protein partitioned into designed side-chains that are
allowed to assume multiple rotamers (red) and a fixed scaffold (blue). (B–D) Different rotamers of the
amino acid tryptophan at a particular position in the protein.

Page 3

Vol. 42 ı No. 1 ı 2007 www.biotechniques.com ı BioTechniques ı 35
Tech Insight
of function is still in its infancy, and successful designs
have taken their cues from biology.
One of the goals of CPD is to produce proteins that bind to
other protein targets with high specificity. Initial investiga-
tions led to the design of protein mutants with increased
specificity for both natural and novel peptide targets. For
example, CPD of calmodulin led to a variant with eight mu-
tations that maintained high affinity for the target peptide,
while showing decreased binding affinity for nontargeted
peptides (28,29). Specificity design was taken a step
further when a PDZ domain mutant was engineered to bind
a new sequence (30).
More recent work on protein-peptide and protein-protein
specificity has revealed the value of negative design.
Negative design is the process by which undesired proper-
ties are considered and designed against. A conceptually
straightforward implementation of negative design is one
in which the scoring function favors mutations that are
stabilizing in the target structure while destabilizing in al-
ternative structures. In the design of coil-coil dimers (31),
four states were considered: (i) the homodimer state; (ii)
the heterodimer state; (iii) the unfolded state; and (iv) the
aggregated state. Successful designs selected amino acids
that were predicted to favor the target dimer state over the
rest. Further evidence of the usefulness of negative design
was demonstrated in the redesign of the colicin E7 DNase-
Im7 immunity protein interface (32).
CPD has shown promise in the engineering of protein-DNA
interactions as well. Currently, selective DNA cleavage is
limited to those sequences for which naturally occurring
specific endonucleases are available. CPD aims to change
that by designing endonuclease variants with altered
specificity (33). The first computational example focused
on mutations that accommodated direct hydrogen bonds
with the altered DNA base pairs. After identifying a suit-
able mutation, the surrounding residues were designed for
optimal side-chain complementarity. This achievement may
pave the way for the automated design of restriction en-
donucleases with therapeutic applications. However, CPD
must first address water-mediated hydrogen bonds before
it can be generally applicable to all DNA sequences. One
possible solution that has been considered is to include
solvated side-chains in the rotamer library (34). The ex-
panded rotamer library contains explicit waters hydrogen-
bonded to polar side-chain atoms. As a result, predictions
can include rotamers that hydrogen bond to DNA through a
bridging water molecule.
The design of protein-ligand interactions poses its own
challenges. Currently, much of the design work on small
ligand binding has been carried out using periplasmic
binding proteins (PBPs) (35). Because wild-type PBPs can
bind to a wide variety of small molecules, they have proven
to be convenient scaffolds. In addition, ligand binding is
easily detected due to the large conformational change as-
sociated with the binding event. Currently, PBPs have been
computationally redesigned to bind metals, neurotransmit-
ters, sugars, explosives, and nerve agents (26,36,37). The
successes in protein-ligand design using PBPs suggest
that scaffold selection is crucial. Focusing the design to a
preexisting binding region on a scaffold shown to accom-
modate a wide variety of ligands appears to improve the
chances of a successful design.
A similar trend seems to be applicable to the design of
catalytic activity. Initial attempts to design an enzyme
that catalyzes the hydrolysis of p-nitrophenyl acetate used
thioredoxin as a scaffold and were modestly successful
(25). The redesign of the binding region of a PBP success-
fully introduced triose phosphate isomerase activity (38).
In both cases, the catalytic residues and transition states
are geometrically constrained in an optimal conformation
for catalysis. A design algorithm then identifies a posi-
tion in the scaffold that accommodates both the catalytic
residues and the transition state. The catalytic binding
pocket is finally repacked for optimal complementarity with
the substrate. The design of catalytic activity in various
scaffolds is an excellent way to extend our understanding
of enzyme catalysis. In addition, the ability to design novel
enzymes that carry out difficult reactions would be of great
use to the pharmaceutical industry.
The recent accomplishments of CPD clearly depict the
field’s ability to address biologically relevant applications.
One of the best examples of this is the computational
Figure 2. Computationally designed tumor necrosis factor (TNF) variant
is depicted in gold. Mutations, highlighted in red, prevent designed variant
from binding to TNF receptors, shown in blue. The addition of the mutated
TNF results in a heterogeneous mixture of trimers, most of which have
compromised activity. Adding an excess of mutated TNF insures low levels
of active native TNF homotrimers. This figure is based on one provided in the
supplementary material of Reference 39. Crystal structure 1TNR was used to
generate the representative structures.

Page 4

Vol. 42 ı No. 1 ı 2007 www.biotechniques.com ı BioTechniques ı 37
Tech Insight
design of a tumor necrosis factor (TNF) variant that is
currently being developed as a therapeutic agent for the
treatment of rheumatoid arthritis. The design focused on
the binding interface between TNF and its receptor. Us-
ing a minimalist approach, mutations that disrupted the
interface between TNF and its receptor were identified
(Figure 2). Studies have shown that the designed variant
inactivates native TNF by sequestering it, using a new
dominant negative mechanism of action (39). The success
of this potential therapeutic highlights the capabilities of
CPD. With further improvements, CPD could revolutionize
protein engineering.
As computational power improves, larger and more so-
phisticated designs will become possible. Furthermore,
the predictive power of rational design is expected to grow
as our knowledge of protein function advances. This will
aid in improving the scoring functions and in identifying
problems with the design model. It is clear from current
results that further conformational sampling of the back-
bone and side-chains would be beneficial. A decrease in
false positives coupled with high-throughput screening will
ultimately lead to a rapid and reliable tool for the develop-
ment of novel proteins with unique functions.
Acknowledgments
O.A. and B.D.A. contributed equally to this work.
References
1. Drexler, K.E. 1981. Molecular engineering—an approach to the
development of general capabilities for molecular manipula-
tion. Proc. Natl. Acad. Sci. USA 78:5275-5278.
2. Pabo, C. 1983. Molecular technology—designing proteins and
peptides. Nature 301:200.
3. Janin, J., S. Wodak, M. Levitt, and B. Maigret. 1978. Conformation
of amino-acid side-chains in proteins. J. Mol. Biol. 125:357-
386.
4. Dunbrack, R.L. and F.E. Cohen. 1997. Bayesian statistical analy-
sis of protein side-chain rotamer preferences. Protein Sci.
6:1661-1681.
5. Dahiyat, B.I. and S.L. Mayo. 1997. Probing the role of pack-
ing specificity in protein design. Proc. Natl. Acad. Sci. USA
94:10172-10177.
6. Dahiyat, B.I., D.B. Gordon, and S.L. Mayo. 1997. Automated de-
sign of the surface positions of protein helices. Protein Sci.
6:1333-1337.
7. Kortemme, T., A.V. Morozov, and D. Baker. 2003. An orientation-
dependent hydrogen bonding potential improves prediction of
specificity and structure for proteins and protein-protein com-
plexes. J. Mol. Biol. 326:1239-1259.
8. Lee, B. and F.M. Richards. 1971. Interpretation of protein struc-
tures—estimation of static accessibility. J. Mol. Biol. 55:379-
400.
9. Street, A.G. and S.L. Mayo. 1998. Pairwise calculation of protein
solvent-accessible surface areas. Fold. Des. 3:253-258.
10. Colonna-Cesari, F. and C. Sander. 1990. Excluded volume
approximation to protein-solvent interaction—the solvent con-
tact model. Biophys. J. 57:1103-1107.
11. Lazaridis, T. and M. Karplus. 1999. Effective energy function
for proteins in solution. Proteins 35:133-152.
12. Zollars, E.S., S.A. Marshall, and S.L. Mayo. 2006. Simple
electrostatic model improves designed protein sequences.
Protein Sci. 15:2014-2018.
13. Kuhlman, B., G. Dantas, G.C. Ireton, G. Varani, B.L. Stoddard,
and D. Baker. 2003. Design of a novel globular protein fold
with atomic-level accuracy. Science 302:1364-1368.
14. Wisz, M.S. and H.W. Hellinga. 2003. An empirical model for
electrostatic interactions in proteins incorporating multiple
geometry-dependent dielectric constants. Proteins 51:360-
377.
15. Pokala, N. and T.M. Handel. 2005. Energy functions for
protein design: adjustment with protein-protein complex af-
finities, models for the unfolded state, and negative design of
solubility and specificity. J. Mol. Biol. 347:203-227.
16. Metropolis, N., A.W. Rosenbluth, M.N. Rosenbluth, A.H. Teller,
and E. Teller. 1953. Equation of state calculations by fast
computing machines. J. Chem. Physiol. 21:1087-1092.
17. Kirkpatrick, S., C.D. Gelatt, and M.P. Vecchi. 1983. Optimiza-
tion by simulated annealing. Science 220:671-680.
18. Voigt, C.A., D.B. Gordon, and S.L. Mayo. 2000. Trading accu-
racy for speed: a quantitative comparison of search algorithms
in protein sequence design. J. Mol. Biol. 299:789-803.
19. Desjarlais, J.R. and T.M. Handel. 1995. De-novo design of the
hydrophobic cores of proteins. Protein Sci. 4:2006-2018.
20. Desmet, J., J. Spriet, and I. Lasters. 2002. Fast and accurate
side-chain topology and energy refinement (FASTER) as
a new method for protein structure optimization. Proteins
48:31-43.
21. Allen, B.D. and S.L. Mayo. 2006. Dramatic performance
enhancements for the FASTER optimization algorithm. J.
Comput. Chem. 27:1071-1075.
22. Desmet, J., M. Demaeyer, B. Hazes, and I. Lasters. 1992. The
dead-end elimination theorem and its use in protein side-
chain positioning. Nature 356:539-542.
23. Gordon, D.B., G.K. Hom, S.L. Mayo, and N.A. Pierce. 2003.
Exact rotamer optimization for protein design. J. Comput.
Chem. 24:232-243.
24. Hellinga, H.W. and F.M. Richards. 1991. Construction of new
ligand binding sites in proteins of known structure. I. Com-
puter-aided modeling of sites with pre-defined geometry. J.
Mol. Biol. 222:763-785.
25. Bolon, D.N. and S.L. Mayo. 2001. Enzyme-like proteins by
computational design. Proc. Natl. Acad. Sci. USA 98:14274-
14279.
26. Looger, L.L., M.A. Dwyer, J.J. Smith, and H.W. Hellinga. 2003.
Computational design of receptor and sensor proteins with
novel functions. Nature 423:185-190.
27. Lassila, J.K., H.K. Privett, B.D. Allen, and S.L. Mayo. 2006.
Combinatorial methods for small-molecule placement in
computational enzyme design. Proc. Natl. Acad. Sci. USA
103:16710-16715.
28. Shifman, J.M. and S.L. Mayo. 2002. Modulating calmodulin
binding specificity through computational protein design. J.
Mol. Biol. 323:417-423.

Page 5

Vol. 42 ı No. 1 ı 2007 www.biotechniques.com ı BioTechniques ı 39
Tech Insight
29. Shifman, J.M. and S.L. Mayo. 2003. Exploring the origins of
binding specificity through the computational redesign of
calmodulin. Proc. Natl. Acad. Sci. USA 100:13274-13279.
30. Reina, J., E. Lacroix, S.D. Hobson, G. Fernandez-Ballester, V.
Rybin, M.S. Schwab, L. Serrano, and C. Gonzalez. 2002. Com-
puter-aided design of a PDZ domain to recognize new target
sequences. Nat. Struct. Biol. 9:621-627.
31. Havranek, J.J. and P.B. Harbury. 2003. Automated design of
specificity in molecular recognition. Nat. Struct. Biol. 10:45-
52.
32. Joachimiak, L.A., T. Kortemme, B.L. Stoddard, and D. Baker.
2006. Computational design of a new hydrogen bond net-
work and at least a 300-fold specificity switch at a protein-
protein interface. J. Mol. Biol. 361:195-208.
33. Ashworth, J., J.J. Havranek, C.M. Duarte, D. Sussman, R.J.
Monnat, Jr., B.L. Stoddard, and D. Baker. 2006. Computa-
tional redesign of endonuclease DNA binding and cleavage
specificity. Nature 441:656-659.
34. Jiang, L., B. Kuhlman, T. Kortemme, and D. Baker. 2005. A
“solvated rotamer” approach to modeling water-mediated hy-
drogen bonds at protein-protein interfaces. Proteins 58:893-
904.
35. Dwyer, M.A. and H.W. Hellinga. 2004. Periplasmic binding
proteins: a versatile superfamily for protein engineering. Curr.
Opin. Struct. Biol. 14:495-504.
36. Marvin, J.S. and H.W. Hellinga. 2001. Conversion of a maltose
receptor into a zinc biosensor by computational design. Proc.
Natl. Acad. Sci. USA 98:4955-4960.
37. Allert, M., S.S. Rizk, L.L. Looger, and H.W. Hellinga. 2004.
Computational design of receptors for an organophosphate
surrogate of the nerve agent soman. Proc. Natl. Acad. Sci.
USA 101:7907-7912.
38. Dwyer, M.A., L.L. Looger, and H.W. Hellinga. 2004. Com-
putational design of a biologically active enzyme. Science
304:1967-1971.
39. Steed, P.M., M.G. Tansey, J. Zalevsky, E.A. Zhukovsky, J.R.
Desjarlais, D.E. Szymkowski, C. Abbott, D. Carmichael, et al.
2003. Inactivation of TNF signaling by rationally designed
dominant-negative TNF variants. Science 301:1895-1898.