PROTEIN ENGINEERING
7. Rational and semi-rational design
Loschmidt Laboratories
Department of Experimental Biology
Masaryk University, Brno
Outline
 Protein engineering approaches
 Semi-rational design
 identification of hot-spots
 evaluation of hot-spots
 selection of substitutions
 design of library
 mutagenesis and screening
 Rational design
 molecular modeling
Outline
 Protein engineering approaches
 Semi-rational design
 identification of hot-spots
 evaluation of hot-spots
 selection of substitutions
 design of library
 mutagenesis and screening
 Rational design
 molecular modeling
Protein engineering
 altering protein structure to improve its properties
 three main approaches
 rational design
 directed evolution
 semi-rational design
Protein engineering approaches
Protein engineering approaches
Rational design Directed evolution Semi-rational design
high-throughput
screening/selection not essential essential
advantageous
but not essential
structural and/or
functional information
both essential neither essential either is sufficient
sequence space
exploration
low high, random moderate, targeted
probability to obtain
synergistic mutations
moderate low high
 worldwide Protein Data Bank (wwPDB)
 http://www.wwpdb.org/
 central repository of ~220,000 experimental macromolecular structures
(April 2024)
 RCSB PDB
 https://www.rcsb.org
 PDBe
 https://www.ebi.ac.uk/pdbe
 PDBj
 https://pdbj.org
Structural information
Structural prediction
 Alpha Fold 2
 Galaxy: https://usegalaxy.eu/?tool_id=alphafold, Colab:
https://colab.research.google.com/github/deepmind/alphafold/blob/
main/notebooks/AlphaFold.ipynb
 structure prediction directly from sequence using deep learning,
evolutionary information (MSA), and structure optimization
 Multimer mode – lower accuracy
 Not precise in sidechain orientations prediction (not appropriate for
protein-ligand interaction - molecular docking)
 Rare folds, alternative conformations, and co-factors not predicted
Structure prediction
 Alpha Fold database
 https://alphafold.ebi.ac.uk
 Database of protein structures predicted by Alpha Fold 2
 Over 200 million sequences modeled (available also in UniProt)
 Protein engineering approaches
 Semi-rational design
 identification of hot-spots
 evaluation of hot-spots
 selection of substitutions
 design of library
 mutagenesis and screening
 Rational design
 molecular modeling
Semi-rational design
 combine advantages of rational and random approaches
 selection of promising target sites (hot-spots) → mutagenesis
→ creation of small “smart” libraries
 based on knowledge of protein structure and function
  high-throughput screening usually not needed
  increased chance of obtaining variants with desired
properties
  certain knowledge of protein structure-function
relationships is still required,  but not that much
 Protein engineering approaches
 Semi-rational design
 identification of hot-spots
 evaluation of hot-spots
 selection of substitutions
 design of library
 mutagenesis and screening
 Rational design
 molecular modeling
Identification of hot-spots
 hot-spots for engineering catalytic properties
 hot-spots for engineering thermostability
Hot-spots for engineering catalytic properties
 residues mediating substrate binding, transition-state
stabilization or product release → mutations can improve or
disrupt binding, catalysis or ligand transport
 residues involved in protein-ligand interactions
 residues located in binding pockets
 residues located in access tunnels
→ these residues also include catalytic or other essential residues which
generally should not be mutated!
Analysis of protein-ligand interactions
 requires 3D structure of protein-ligand complex
 experimental structure (wwPDB, PDBbind)
 theoretical model (molecular docking)
Analysis of protein-ligand interactions
 schematic diagrams of protein-ligand interactions
LigPlot, LigPlot+ PoseView
Analysis of binding pockets
 binding and active sites of enzymes are often associated
with structural pockets and cavities
ligandactive site
pocket
Analysis of binding pockets
 binding and active sites of enzymes are often associated
with structural pockets and cavities
Analysis of binding pockets
 binding and active sites of enzymes are often associated
with structural pockets and cavities
mutation
Analysis of binding pockets
 binding and active sites of enzymes are often associated
with structural pockets and cavities
Analysis of binding pockets
 binding and active sites of enzymes are often associated
with structural pockets and cavities
Analysis of binding pockets
 binding and active sites of enzymes are often associated
with structural pockets and cavities
 most amino acid residues located in these pockets may come into
contact with the ligands during the catalytic cycle
→ one can accurately predict which residues may interact with the
ligand even without precise knowledge of ligand orientation in the
active site
 requires 3D structure of protein
 software for detection of pockets
 CASTp, fPocket, CavityPlus, etc.
Analysis of binding pockets
 detailed characterization of all pockets in the structure
CavityPlus
Analysis of access tunnels
 buried binding or active sites are connected with bulk
solvent by access tunnels
ligand
active site
pocket
access tunnel
Analysis of access tunnels
 buried binding or active sites are connected with bulk
solvent by access tunnels
Analysis of access tunnels
 buried binding or active sites are connected with bulk
solvent by access tunnels
mutation
Analysis of access tunnels
 buried binding or active sites are connected with bulk
solvent by access tunnels
mutation
Analysis of access tunnels
 buried binding or active sites are connected with bulk
solvent by access tunnels
 adjusted to permit transport of specific molecules
 mutations can speed-up or hinder transport of molecules as well
as allow transport of other molecules
 requires 3D structure of protein
 software for detection of tunnels
 Caver, Mole, HOLE, PoreWalker
Analysis of access tunnels
 Caver Web
 https://loschmidt.chemi.muni.cz/caverweb
Radius[Å]
Length [Å]
Hot-spots for engineering thermostability
 highly flexible residues – introduction of rigidifying mutations
 residues located in access tunnels
→ these residues may also include catalytic or other essential
residues which generally should not be mutated!
Identification of highly flexible residues
 prediction based on crystallographic B-factors
 reflect the degree of thermal motion, and thus the flexibility of
individual residues
 requires 3D structure of protein
 experimental structure determined by X-ray crystallography (wwPDB)
Identification of highly flexible residues
 average B-factor of each residue in the target protein
B-FITTER
Analysis of access tunnels
 saturation mutagenesis in tunnel residues has 2× better
chance to significantly improve stability than mutagenesis in
other protein regions (based on computational predictions)
 Protein engineering approaches
 Semi-rational design
 identification of hot-spots
 evaluation of hot-spots
 selection of substitutions
 design of library
 mutagenesis and screening
 Rational design
 molecular modeling
Evaluation of hot-spots
 hot-spots identified by computational tools can be further
evaluated to prevent replacing indispensable amino acid
residues and to prioritize the hot-spots (i.e., order the hotspots
based on their suitability for mutagenesis)
 analysis of evolutionary conservation
 prediction of effects of mutations on protein stability or
function
Analysis of evolutionary conservation
 residues essential for maintaining structural or functional
properties of a protein tend to be conserved during evolution
 conserved residues are generally not recommended as suitable
targets for mutagenesis - their replacement often leads to the loss of
protein function
 mutagenesis targeting highly mutable positions provides a significantly
higher proportion of viable variants than random mutagenesis
 targeting moderately or highly variable positions, which are expected
to be tolerant to a wide range of substitutions, represents a good
approach for producing efficient smart libraries (i.e., libraries
with a high proportion of correctly folded and active variants)
Analysis of evolutionary conservation
 residue conservation can be derived from a multiple
alignment of a set of related proteins (3D structure not
required)
Analysis of evolutionary conservation
 residue conservation can be derived from a multiple
alignment of a set of related proteins (3D structure not
required)
Analysis of evolutionary conservation
 evolutionary conservation of individual positions in protein
mapped on protein 3D structure
ConSurf
Prediction of mutation effects
 computational tools for the prediction of effect of amino
acid substitutions on protein stability or protein function
 in silico site-saturation mutagenesis of identified hot-spots – check if
mutations at a given site are likely to be tolerated
 many highly destabilizing/deleterious mutations predicted for a
certain position – given site is not a very good target for
mutagenesis
 sites with only a few highly destabilizing /deleterious mutations
predicted can still represent promising hot-spots (the amino acids
with potentially destabilizing/deleterious effects can be discarded
from the library by the appropriate selection of degenerate codons)
Prediction of mutation effects
 effects on protein stability – usually requires 3D structure
of protein
 experimental structure (wwPDB)
 theoretical model (AlphaFold, homology modeling)
 effects on protein function – sequence information often
sufficient
Prediction of mutation effects
 prediction of effect of substitutions on protein stability
 Evaluation of the change of protein free energy upon mutation
 Evaluation of contributions of individual interactions to total
energy
 Usually requires structural information
 software for prediction of effect of mutation on stability
 Rosetta, FoldX, CUPSAT, ERIS
Prediction of mutation effects
 prediction of effect of substitutions on protein stability
Rosetta in HotSpot Wizard
Prediction of mutation effects
 prediction of effect of substitutions on protein function
 Evaluation if a mutation would impair protein function
 Hard to describe by physico-chemical properties > machine learning
 Usually sequence based calculation
 software for prediction of effect of mutation on function
 PredictSNP, SIFT, MAPP, PhD-SNP…
Prediction of mutation effects
 prediction of effect of substitutions on protein function
PredictSNP
 AlphaMissense
 https://github.com/google-deepmind/alphamissense
 deep learning predictor based on AlphaFold
 analysis of human and primate missense mutations
 trained on population frequency data and uses sequence and
predicted structural context
 all single–amino acid substitutions in the human proteome are
provided
Prediction of mutation effects
 Protein engineering approaches
 Semi-rational design
 identification of hot-spots
 evaluation of hot-spots
 selection of substitutions
 design of library
 mutagenesis and screening
 Rational design
 molecular modeling
Selection of substitutions
 substitutions introduced using degenerate codons
 e.g., NNK (N = A/T/G/C; K = T/G)
Selection of substitutions
 all possible substitutions - NNK or NNS degenerate codons
  encode all 20 amino acids with the lowest redundancy and price
(mixture of 32 codons)
  redundancy is not completely eliminated (3× Arg, Leu, Ser, 2×
Ala, Gly, Pro, Thr and Val)
Selection of substitutions
 all possible substitutions - NNK or NNS degenerate codons
 introduction of only selected substitutions using
degenerate codons encoding reduced amino acid alphabets
  do not encode all 20 amino acids
  decreased library size → improved screening efficiency
 NDT – balanced set of 12 amino acids (12 codons)
Selection of substitutions
 all possible substitutions - NNK or NNS degenerate codons
 introduction of only selected substitutions using
degenerate codons encoding reduced amino acid alphabets
Selection of reduced amino acid alphabets
 introduction of amino acids exhibiting certain properties
 VRK – 8 hydrophilic amino acids (12 codons)
 NYC – 8 hydrophobic amino acids (8 codons)
 KST – 4 small amino acids (4 codons)
 ...
Selection of reduced amino acid alphabets
 introduction of amino acids exhibiting certain properties
 introduction of a balanced set of amino acids
 NDT – balanced set of 12 amino acids (12 codons)
Selection of reduced amino acid alphabets
 introduction of amino acids exhibiting certain properties
 introduction of a balanced set of amino acids
 introduction of substitutions existing (at a given site) in
known natural proteins
 likely increasing the proportion of viable variants in the resulting
library
 can be obtained by analysis of multiple sequence alignment
Selection of reduced amino acid alphabets
 introduction of amino acids exhibiting certain properties
 introduction of a balanced set of amino acids
 introduction of substitutions existing (at a given site) in
known natural proteins
 discarding amino acids with potentially destabilizing/
deleterious effects
 can be obtained by prediction of effects of mutations on protein
stability or function
HotSpot Wizard
 meta-server combining several tools
 automatic identification of hot-spots for engineering of enzyme
catalytic properties
 prioritization of hot-spots by their mutability
 distribution of amino acids at individual positions
 prediction of stability
 molecular docking
 design of smart libraries
HotSpot Wizard
Functional hot-spots Stability hot-spots (flexibility)
Stability hot-spots (evolution)
P
P
P
P
P
P
P
P
P
P
P
P
P P P
W
W
- - - -
G
G
G
G
G
G
G
G
G
G
GP
P
P
P
P
P
P
P
P
P
R
R
R
R
R
R
R
R
R
R
R
M
M
M
M
M
M
M
M
M
M
M
I
I
I
I
I
I
I
I
I
I
I
T
T
T
T
T
T
T
T
T
T
Y
Y
Y
Y
Y
Y
Y
Y
Y
A
A
A
A
A
A
A
A
A
A
A
S
S
S
S
S
S
S
S
S
S
W
W
W
W
W
W
W
W
W
W
W
W
C
C
C
C
C
C
C
C
H
H
H
H
H
H
H
H
S
S
S
S
S
S
S
S
S
S
S
L
L
L
L
L
L
L
L
L
L
L
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
L
L
L
L
L
L
L
L
L
L
LI
I
L
L
Y
Y
Y R
Correlated hot-spots
P
P
P
P
P
P
P
P
P
P
P
P
P P P
W
W
- - - -
G
G
G
G
G
G
G
G
G
G
GP
P
P
P
P
P
P
P
P
P
R
R
R
R
R
R
R
R
R
R
R
M
M
M
M
M
M
M
M
M
M
M
I
I
I
I
I
I
I
I
I
I
I
T
T
T
T
T
T
T
T
T
T
Y
Y
Y
Y
Y
Y
Y
Y
Y
A
A
A
A
A
A
A
A
A
A
A
S
S
S
S
S
S
S
S
S
S
W
W
W
W
W
W
W
W
W
W
W
W
C
C
C
C
C
C
C
C
H
H
H
H
H
H
H
H
S
S
S
S
S
S
S
S
S
S
S
L
L
L
L
L
L
L
L
L
L
L
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
L
L
L
L
L
L
L
L
L
L
LI
I
L
L
Y
Y
W
W
W
R
R
R
K
K
K
K
K
W
W
W
D
D
D
E
E
E
E
E
V
V
V
HotSpot Wizard
1. protein structure
HotSpot Wizard
2. residues indispensable for protein function: catalytic and
binding residues
HotSpot Wizard
3. functional residues: active site pocket and tunnels
HotSpot Wizard
4. mutability of individual positions of protein
HotSpot Wizard
Hot spots
Tunnels
Cavities
Docking
Stability
Design
Library
HotSpot Wizard
 Protein engineering approaches
 Semi-rational design
 identification of hot-spots
 evaluation of hot-spots
 selection of substitutions
 design of library
 mutagenesis and screening
 Rational design
 molecular modeling
Design of library
 decisions to be made after evaluation and prioritization of
hot-spots:
 how many and which positions to target?
 should the positions be randomized simultaneously or separately?
 should all or only a reduced set of amino acids be introduced at
individual positions?
→ dramatic effect on the size of the resulting library
Design of library – HotSpot Wizard
Design of library – HotSpot Wizard
 Protein engineering approaches
 Semi-rational design
 identification of hot-spots
 evaluation of hot-spots
 selection of substitutions
 design of library
 mutagenesis and screening
 Rational design
 molecular modeling
Mutagenesis and screening
 saturation mutagenesis - next lecture 
 Protein engineering approaches
 Semi-rational design
 identification of hot-spots
 evaluation of hot-spots
 selection of substitutions
 design of library
 mutagenesis and screening
 Rational design
 molecular modeling→ design of mutations
Rational design
 site-specific changes on the target enzyme
 few amino-acid substitutions that are predicted to elicit
desired improvements of enzyme function
 based on detailed knowledge of protein structure, function
and catalytic mechanism
  relatively simple characterization of constructed variants
  complexity of protein structure-function relationships
  molecular modeling expertise usually required
 Protein engineering approaches
 Semi-rational design
 identification of hot-spots
 evaluation of hot-spots
 selection of substitutions
 design of library
 mutagenesis and screening
 Rational design
 molecular modeling→ design of mutations
Molecular modeling
 “Theoretical or computational technique that provides insight
into the behavior of molecular system.”
A. R. Leach
 Ligand binding
 Molecular docking
 Protein dynamics and transport of molecules
 Molecular dynamics
 Reaction barriers and mechanisms
 Quantum chemistry or QM/MM
 Protein design
 Molecular mechanics, machine learning
Molecular docking
 predicts structure of receptor (protein) – ligand complex
Molecular docking
 Two components procedure
 searching – finding the
conformation of ligand in the
active site of the enzyme
 scoring – evaluation of the
binding free energy
 Docking software
 Autodock, Vina, Gold,
Medusa, Rosetta Dock…
Molecular docking
 Virtual screening
 database of compounds + protein structure > molecular docking >
re-scoring > compound prioritization > experimental testing
 Principle
 physical description of interactions within the system (force field)
 Newton’s laws of motions
 forces acting on all atoms due to all atoms
 Provides information on energetics, amplitudes, and time
scales of local motions on the atomic level
Molecular dynamics
force field
Analysis of interactions
Molecular dynamics
Ligand transport
Interaction with membraneLigand conversion
Quantum chemistry
 Modeling of reaction barriers
 Enzymes increase speed of chemical reactions by decreasing
activation barrier
Quantum chemistry
 Using quantum mechanics to create or break bonds (usually
hybrid quantum mechanics/ molecular mechanics simulation)
Design of stability
 FireProt
 https://loschmidt.chemi.muni.cz/fireprotweb
 In silico analysis of all mutations
 Energy- and evolution-based analyses
 Multiple-point mutants for gene synthesis
 Single-point prediction
 User-defined mutations
Design of stability
 FireProt
 FireProtASR
 https://loschmidt.chemi.muni.cz/fireprotasr
 ancestral sequence reconstruction
 Analysis of protein evolution, sequence-based protein stabilization
 Ancestrals are highly stable, have broad specificity and good yields
Design of stability
 FireProtASR
Design of stability
Design of solubility
 ProteinMPNN
 https://huggingface.co/spaces/simonduerr/ProteinMPNN
 deep learning model for protein optimization via mutations
 takes structure on the input and provides optimized sequence
folding into the same backbone
 good for improving yields and rescuing folding-compromised designs
 AffiLib
 https://affilib.weizmann.ac.il/bin/steps
 RosettaDesign and evolution analysis to optimize macromolecular interface
 mutations for improvement of the binding affinity
 up to 50 multiple-point mutants for protein synthesis
Design of protein-protein interactions
Design of mutations
MD of free protein
Docking
MD of complex
QM of complex
Protein design
Experiment Investigated system
Knowledge
Design of mutations
 design of modified enzymes by in silico screening
 study of effects of all relevant mutations
 selection and combination of the best mutations
PROTEIN ENGINEERING
8. Directed evolution
Loschmidt Laboratories
Department of Experimental Biology
Masaryk University, Brno