protein modeling Flashcards
Levels of Protein Organization
Prmary , Secondary , tetriary, Quaternary structure
Proteins intramolecular interactions
hydrogen bonds , electrostatic attractions , van der waals attractions , hydrphobic effect
Spectroscopy methods for structural biology ?
x ray crystallography
Nuclear magnetic resonance
Cryo-Electron Microscopy
All these methods require a large quantity of pure protein
Proteins expression methods
Extracted directly from conventional cells
Bacterial expression
Free of cell expression
yeast cells expression
Insect and Human embryonic kidney (HEK) cells expression
Proteins purification ?
- Cells lysis
- Centrifugation to get
the fraction of interest
For membrane proteins
more steps are required
multiple purification processes Purification 1,2… or more
Assess protein
purity with SDS
page Gel
protein purification methods ?
Size exclusion chromatography
affinity chromatography
Nuclear magnetic resonance (NMR)
Multidimentional NMR
Nuclear magnetic resonance (NMR)
advantages and disadvantages ?
Advantages:
-Can probe dynamic and intrinsically disordered proteins
-Non-destructive and non-invasive
-Three-dimensional structures in their natural state can be measured directly in solution
- Require less protein than X-ray crystallography
Disadvantages:
-Limited to small proteins because of the difficulty of interpretation of biomolecules with large molecular weight proteins
y
how it works
. In presence of a magnetic field there are two spin states with a high and a low energy state
. It is possible to change the nucleus state by application of a second magnetic field
. The change of state upon relaxation (return to the initial state) is mesured to get NMR chromatograms
The chemical shift in NMR is extremely important, as it gives information about the local structure surrounding the nucleus of interest.
X-ray crystallography
Advantages:
-Useful for large structures: Not limited by size or atomic weight
-Provide Higher average resolution than Cryo-EM
-Relatively simple to use when conditions are optimized
-Adapted to small molecules characterization
Disadvantages:
-Require a massive quantity of proteins
-Challenging for membrane proteins
-The crystal must diffract to high resolution
-Conformation can be altered due to the crystal organization
-Can not probe proteins’ dynamic
-Phase Problem
the crystalisation is a n empyrical proccess
Fourier Transform isa mathematical function that helps to transform the signals between two different domains, such as in this case, transforming the signal from frequency (1/d) domain to distance (d) domain.
X-ray crystallography Phase problem
Molecular replacement uses phase data from a similar protein
Isomorphous replacement uses heavy atoms properties to deduce the phase
Direct Method Based on structure factor magnitude of different Fourier components limited to molecules with ∼2000 nonhydrogen atoms and data at high resolution, <1.2 Å
Phase information is necessary for FT calculation and thus to solve protein structure
in detail:
X-ray Diffraction: In X-ray crystallography, a crystal of the sample of interest is irradiated with a beam of X-rays. The X-rays interact with the electron density in the crystal, resulting in constructive and destructive interference patterns, known as diffraction spots, that are captured on a detector.
Diffraction Pattern: The diffraction pattern consists of spots whose positions and intensities are related to the structure of the crystal. The positions of the spots provide information about the spatial arrangement of atoms in the crystal (i.e., the structure), while the intensities of the spots are related to the electron density of the crystal.
Phases: The diffraction pattern contains information about both the amplitude and phase of the scattered X-rays. The amplitude is related to the intensity of the diffraction spots, while the phase is related to the position of the waves within each spot.
Phase Problem: While the amplitudes of the scattered X-rays can be measured directly from the diffraction pattern, the phases cannot be directly determined. This is because the intensity measurements provide only the squared amplitude of the complex wavefunction, which does not contain phase information.
Solving the Phase Problem: Solving the phase problem involves determining the phases of the scattered waves in order to reconstruct the electron density map of the crystal. Several methods have been developed to solve the phase problem, including:
Direct methods, which use mathematical relationships between the phases and amplitudes of the diffraction data to calculate the phases directly.
Molecular replacement, which involves using a known model or structure similar to the unknown structure as a starting point to solve for the phases.
Experimental phasing techniques, such as multiple isomorphous replacement (MIR) or single-wavelength anomalous dispersion (SAD), which involve introducing heavy atoms or exploiting anomalous scattering to obtain phase information.
Iterative Refinement: Once initial phases have been determined, the electron density map is calculated and refined iteratively to improve the accuracy of the structure. This process involves comparing the calculated diffraction pattern with the observed diffraction data and adjusting the phases and atomic coordinates until a model consistent with the experimental data is obtained.
Cryo-Electron Microscopy advantages and disadvantages ?
Advantages:
-Useful for large structures: Not limited by large proteins
-Proteins are imaged in a native like state
-Can deal with a limited amount of flexibility
-Require less quantity of proteins than X-ray crystallography
Disadvantages:
-Can deal with a limited amount of flexibility
-Size limitation for proteins smaller than 60 kDa
-The protein needs to behave well on CryoEM support
(avoid aggregation or preferential orientation)
Cryo-Electron Microscopy how does it work ?
Purified protein
Freezing / Negative
staining
EM data
collection
Particle picking
Particle alignment
and classification
3D model
reconstruction
Model refinement
Global refinement with constraints ?
Global Optimization: Global refinement with constraints involves optimizing the atomic model while simultaneously satisfying all imposed constraints. This global optimization process ensures that the refined model is physically realistic, chemically plausible, and consistent with experimental data.
done with rosseta commons , CCP4, phenix
Manual inspection also advised
Model quality is limited by the resolution of the map, what are the possible resolutions?
4.8Å resolution map
(no side chain informations)
3.5Å resolution map
(partial side chain informations)
2.5 Å resolution map
(side chain informations)
And check local resolution, Maps are not homogeneous!!!
PDB and mmCIF atimic coordinates ?
Both the PDB (Protein Data Bank) and mmCIF (macromolecular Crystallographic Information File) formats are widely used formats for storing structural data of biological macromolecules, such as proteins, nucleic acids, and complexes. These formats contain atomic coordinates that define the positions of atoms within the molecule’s three-dimensional structure. Here’s how atomic coordinates are represented in each format:
PDB Format:
In the PDB format, atomic coordinates are typically represented as a list of lines, each corresponding to a single atom in the molecule.
Each line begins with the keyword “ATOM” or “HETATM” (for atoms belonging to the main chain or to hetero groups, respectively), followed by fields specifying the atom’s serial number, atom name, residue name and number, chain identifier, coordinates (x, y, z), occupancy, temperature factor (B-factor), and optional additional fields.
The atomic coordinates are typically stored in Angstrom units (1 Å = 0.1 nm) and represent the positions of atoms in a Cartesian coordinate system.
Example of an ATOM line in PDB format:
mathematica
Copy code
ATOM 1 N ASP A 1 1.876 5.512 -9.364 1.00 0.00 N
mmCIF Format:
In the mmCIF format, atomic coordinates are represented using a structured data format based on key-value pairs and data blocks.
Atomic coordinates are stored within the data blocks corresponding to individual structural models or entries.
The mmCIF format includes specific data categories and fields for representing atomic coordinates, such as “_atom_site.label_atom_id”, “_atom_site.Cartn_x”, “_atom_site.Cartn_y”, and “_atom_site.Cartn_z”, which specify the atom names and Cartesian coordinates of atoms, respectively.
Similar to the PDB format, the coordinates are typically stored in Angstrom units.
Example of atomic coordinates in mmCIF format:
mathematica
Copy code
_atom_site.label_atom_id _atom_site.Cartn_x _atom_site.Cartn_y _atom_site.Cartn_z
N 1.876 5.512 -9.364
In both formats, the atomic coordinates provide essential information about the spatial arrangement of atoms within a molecule’s structure, enabling visualization, analysis, and manipulation of the molecular model using computational tools and software.
Small moleculescanbedisplayedin different ways , how do we make the computer understand ?
CNCC@Hc1ccc(O)c(O)c1
Adrenaline
how does The information content increases with„dimensions?
„1D“→
Elements and their occurrence
→
Can contain someinformation about atom connectivity
→
Example: SMILES
*
“2D”→
1D + some information on spatial orientation
→
Example: SDF
*
“3D”→
2D + more detailed information on spatial orientation
→
Example: PDB, MOL2
SMILES strings, what are they ?
*
simplified molecular-input line-entry system
*
Line representation of chemical structure of a molecule
*
Basic information on atom connectivity
*
Usually there are different ways to correctly write the same molecule
→
E.g. ethanol: CCO, OCC and C(O)C
SD file ?
*
SDF = structure-data file
*
Someinformationon spatialorientation
*
Associatesdata with structure
*
Molecules display similar toMOL file
SD File (or SDF: Structure Data File):
SD files are more versatile and can store both 2D and 3D chemical structure information for small molecules.
SD files are often used to store additional data beyond basic structure information, such as compound names, identifiers, properties, and experimental data.
SD files support a tabular format, allowing multiple compounds and associated data to be stored in a single file, with each compound represented by a block of data.
SD files can include both text and binary data, making them suitable for storing a wide range of chemical and biological data.
SD files are commonly used in cheminformatics applications, chemical databases, and virtual screening studies.
Mol2 file ?
*
Spatialinformationofmolecule
*
Information on atomtypes
*
Information on connectivity
*
Sometimesinformationon partial charges
MOL File (or Molfile):
MOL files are simple and widely used text-based file formats for representing the 2D chemical structure of small molecules.
A MOL file typically contains information such as atom connectivity, bond orders, atom coordinates, and atom types.
MOL files are commonly used for storing and exchanging chemical structure data between software applications and databases.
MOL files can represent a single molecule or a collection of molecules in a single file, with each molecule separated by a “$$$$” delimiter.
MOL files are human-readable and can be easily edited with a text editor.
Why search in molecule/ligand databases?
*
You want to know….
→
…characteristics of a specific molecule
→
…whether a molecule binds to a protein
→
…whether ligands for a protein are known
→
…whether a novel ligand you discovered is similar toknown ligands
→
…whether your docking setup works
Thereareseveraldatabaseswithdifferent information
*
Whatinformationareyousearchingfor?
*
Whatinformationisavailablein thespecificdatabase?
What to do if there is no experimental structure?
*
Only few residues difference?
→
In silico point mutations might be sufficient
*
Structure of a protein with a similar sequence available?
→
Homology Modelling
*
No good templates? No specific details required?
→
Ab initio folding
what is the basis of Homology modelling ?
*
“Classical” approach to create structural models
*
Comparative approach
*
“Similar sequences have a similar 3D structure”
Homology modelling Why Choosing a good template is essential?
Choose the right template
→
An experimental proteinstructureofa similarprotein
→
The more similar the sequence, the better the prediction!
→
Sequence identity: at least 25-30% for decent results
→
Sequence similarity
→
How good is the resolution of the template?
→
Is there more than one template?
→
Is the entire target sequence covered? Loop modelling required?
→
Any additional considerations? E.g. specific conformations?
what is the difference between sequence identity and sequence similarity
?
Sequence identity refers to the exact matching of residues (nucleotides or amino acids) at the same position in aligned sequences.
Sequence similarity refers to the degree of likeness or resemblance between two sequences, taking into account not only identical residues but also similar residues.
what to consider ? Homology modelling,
Carefully align similar regions
2.
Align the sequence of target and template(s)
→
Which parts of the sequence are similar?
→
Are there parts you don’t want to align?
what should you consider ? Homology modelling,
Extracting information from the template
Extracting spatial restraints
→
How is the spatial environment of a residue in the template?
→
Transfer the information to the target
Homology modelling,
what should you consider while
Modelling?
→
Transfer the spatial orientation from the template to the structure
→
Keep as many restraints as possible
Homology modelling, What should you consider during Refinement to improve the model ?
Refine the model
→
Side chain orientations
→
Removing clashes
→
Energy minimisation
→
….
Homology modelling what should you consider while evaluating the model ?
Don’t just use, what you get
Evaluation
→
Checking for clashes
→
Weird side chain orientations
→
Ramachandran plot
→
Kinked backbones
→
….
what is the Ramachandran plot ?
*
Does the backbone adopt angles and conformations that are theoretically allowed?
Homology modelling ALL the steps ?
1.
Choose the right template
2.
Align the sequence
3.
Extracting spatial restraints
4.
Modelling
5.
Refine the model
6.
Evaluation
if you dont like what you see you start again with different input
Homology modelling
When does it work well?
- Template and Target have a high sequence similarity/identity
- Known structural motifs
→ e.g. α-helices
Homology modelling
When is it less accurate?
- Loops (structurally more complex)
- No templates with similar sequence
- Reproducing different protein conformations ( Choice of template improtant )
- Details e.g. binding site conformation & side chain orientations ( you can model with a ligand in the binding site)
Homology modellingSummary: What to watch out for…
*
Choose templates carefully
→
Resolution
→
Sequence similarity
→
Conformation
→
Bound ligand?
*
Pay attention to how the sequences are aligned
*
Evaluate a model before using it further
*
Is an additional optimization required?
Ab initio folding, what is it ?
*
“Predicting the 3D structure only based on the amino acid sequence”
*
Not a new concept but incredibly challenging
*
Massive advances in the recent years by using AI/ML approaches
→
AlphaFold2
→
RosettaFold
→
OmegaFold
→
OpenFold
→
ESMFold
Code available online
Ab initio foldingUsing AlphaFold2 , how does it work ?
*
AlphaFolddatabase: pre-predicted structures
*
ColabFold:
→
Colabnotebook for custom protein structure prediction
*
Local installation:
→
Make your own predictions (probably doesn’t work on the average PC)
*
Available code:
→
Can be downloaded and adapted
*
Adaptations of the original code:
→
AlphaFoldMultimer: Protein complex predictions
→
Prediction of different conformations
Ab initio folding
how does AlphaFold2 prediction confidence work ?
*
Confidence metric: pLDDT(based on Local Distance Difference Test metric)
*
pLDDT>90: expected to be modelled with high accuracy
*
90>pLDDT>70: well modelled, generally good backbone prediction
*
70>pLDDT>50: low confidence modelling, be cautious
*
50>pLDDT: should not be interpreted, likely disordered (unstructured or only structured in complex with other protein
Ab initio foldingAlphaFold2 –problems and difficulties
*
Higher confidence for structured domains with many inter-residue contacts
*
Low confidence for loops, linkers & unstructured regions
*
Ignorant of different conformations (in parts fixed by adapted, separate codes)
*
No predictions for non-protein components
*
Lack of details (e.g.binding sites) & no way to directly influence this
*
Ignorant of environment, e.g.membra
The value of structure prediction techniques ?
*
Reasons to use a predicted structure:
→ No experimental structure available
→ No experimental structure in the desired conformation available
→ No full-length experimental structure available
*
Homology modelling?
→ Modelling of specific features in a specific way (e.g. the ligand binding site)
→ More influence on specific features and how they should be modelled
*
AI-generated models?
→ No idea about the structure at all
→ No templates available
→ Quick impression of structural arrangement
→
No details required (e.g. as input for MD)
how do we assess the Similarity of structures?
- Root-mean-square deviation of atomic positions (RMSD)
- Quantitative measure for the similarity of two protein structures of the same protein
- Usually calculated for the protein backbone (C, O, N and Cα) or Cα only
- Usually also includes a rigid superimposition to minimize the resulting RMSD
in vitro screen vs in silico screen ?
in vitro
Requires more time to cover
a small portion of space
More definite yes or no
in silico
Missing a few good choices
Requires less time to cover
a large portion of space
MethodsofComputer-AidedDrug Design (CADD)
STRUCTURE BASED
*
Docking calculations
*
Virtual screening
*
MD simulations
*
…
Keywords:
*
Homology Modelling
*
Force Fields (Molecular Mechanics)
Ligand-based
*
Pharmacophore modelling
*
MedChemapproach
…
Docking calculations
How do twomolecules interact?
Prediction of the interactions between two molecules
* Protein – Small molecules
* Protein – Protein
tses 2 have same basics but different methodology
* (Small molecule – Small molecule)
The steps of a virtual docking screen ?
Protein structure, Molecule library
Docking calculation
Ranking
Post-processing (clustering, filtering, …)
Visual inspection & molecule selection
How are we Preparing the protein structure?
Are there artifacts from exp. structure (stabilization)?
*
Correct mutations
*
Remove additional proteins
Molecules from crystallization? (e.g.PEG)
*
Remove
Additional molecules, proteins or protein parts (e.g. fusion proteins) can slow down the calculation!
Are side chains missing or with two orientations?
*
Add missing side chains
*
(Rotamer libraries to predict orientation)
Hydrogen atoms
*
Add if they are missing
*
Consider protonation states!
→Can depend on the environment!
tricky… where to protonate histidine for example
Water molecules in the binding site?
*
Mediators of protein-ligand interactions?
Choseyour molecule set wiselyWhat’s the goal?
Novel ligands
*
Diverse chemotypes
*
Molecules with diverse characteristics(MW, logP, charges, H-bond donors/acceptors..)
*
(Ultra-)Large libraries
*
virtual molecule libraries grew drastically in recent years!
Specific ligands
*
Based on prior knowledge
*
In-house libraries
*
Natural products
hw do we Prepare the molecules for docking?
*
Add hydrogen atoms
*
Correct protonation states?
→
pH?
*
Usually 3D-conformations are necessary
→
Depending on docking program
→
Conformer generation can be tricky
how is the docking step done ?
In principle docking consists of two repeating steps
Search for poses (Orienting the molecule in the binding site)
Scoring
(Judging the pose)
what are the 2 things to consider Searching for molecule poses ?
- What is kept rigid?
→
Protein & ligand (rarely used nowadays)
→
Protein
→
Neither protein nor ligand (more complex) - Search algorithm
→
Stochastic (e.g.genetic algorithms)
→
Deterministic (e.g.Energy minimization)
→
Systematic (e.g.using an ensemble of pre-created conformers)
what are the Force Fields based on Molecular Mechanics that we use in Scoring the molecule pose ?
Bond
Angle
Torsion
but mainly these cause the other ones dont change easily :
Electrostatic interactions
Van-der-Waals interactions
There are different types of scoring functions , some of them are…
Scoring functions:
Force field based
Molecular mechanics
Empirical
Reproduce empirical data of a specific system
(statistical analyses of experimental data)
Knowledge based
Some parameters stem from empirical data
Another important factor:
Entropic contributions (e.g.desolvation, restraining of flexibility,…)
Partially included in some scoring functions; very complex to model!
In general: The lower the energy value, the better!
Ranking the molecules, how is it done ?
The more exact the Scoring function, the better the prediction!
→More time intensive
→What is your aim?
→Few molecules & exact poses:
more accurate, but slower scoring function
→Many molecules, quick poses over accuracy:
less accurate, but faster scoring function
Good rank ≠ True ligand
Scoring functions are not perfect…
Visual evaluation:Trying to balance deficits of the scoring function , how is it done ?
UsuallyTop 500-5000 molecules from the ranking
Intramolecular angles?
Clashes?
Are polar groups interacting?
Desolvation; stronger interactions
Apolarinteractions?
e.g.π-πinteractions
Docking calculations have their limits, which are they ?
- Predicting ligand binding poses
- Differentiate between ligands and non-binders
- Discovery of chemically novel ligands
- Predict absolute affinities
- Differentiate between ligands and non-binders
- Positioning of very flexible ligands (e.g. peptides)
- Considering protein flexibility
Which questions can be answered with docking?
Screen large molecular libraries
Virtual Screening
Rational modification of a hit molecule
Rational combination of fragments
Which amino acids are relevant for Protein-Ligand interaction?
Don‘t forget about the experimental validation!
Does the GPCR conformation influence which hits we get?
YES
EXAMPLE
Inactive conformation
favours antagonists
Active conformation
favours agonists
Finding antagonists with novel scaffolds
ZINC12 library
~3.6 mio. mols
27 molecules
tested
1 sub-micromolar hit=Novel scaffold!
Create a library of
similar molecules
Docking &
evaluation
All 11 additionally
tested molecules
are ligands!
When can we consider a docking screen successful?
What was the aim?
* Screening for any ligands to an “easy” target? → higher hit rate (30-40%?)
* Searching for novel scaffolds? → lower hit rate might be fine
* Screening to a new binding site, a target without ligands or a model? → 1 hit might be ok
(Virtual) screens are not supposed to yield the perfect new drug, but only a starting point!
Why simulations?
To study details that
are not easily
measured in real-life
experiments.
To study small and fast
phenomena.
Biomolecular
simulations are most
commonly “classical
MD”
We simulate proteins to
validate and quantify the
function, motions and
interactions that determine
molecular function.
Computation electrophysiology
Lead optimization
(Generate the hit analogues by linking
the hit with fragment molecules.) and testing them
How are simulations carried out?
Classical” MD = Newtons laws of motion
a small change in one atom can make a huge difference in many after a few steps ,
1) Position + Velocity of all atoms
2) Interactions with all other atoms
(forcefield)
3) Calculate the forces acting on all
atoms
4) Update atoms to new positions
and velocities after a small time
step.
5) Go to step 1 and repeat
simulations can involve structure up to the whole cell with lees resolution as we go up in size
How small is our timestep? ( simulations )
2 fs (0.000 000 000 000 002s)
How many atoms to simulate this protein?
12 000 (144M interactions)
How much time could you simulate of this
protein on a modern laptop in a day?
10 ns (5M steps)
KTH supercomputer:
256’000 cores to be able to proccess more
Now in practise,Grab a protein from the pdb and start simulating? If not what do we need to consider ?
non-protein molecules (crystallographic waters,
ligands, modified amino acids, etc.)
● alternate conformations
● missing side-chain atoms
● missing fragments
● clashes between atoms
● multiple copies of the same protein chains
● di-sulfide bonds
● wrong assignment of the N and O atoms in the
amide groups of ASN and GLN, and the N and C
atoms in the imidazole ring of HIS [
7Building (preparing) a structure for simulation , how is it done in 7 steps ?
- Clean
- Box
- Solvate ( add water )
- Neutralize
(add ions) - Minimize
energy - Equilibrate 7. Production MD
and analysis
what are some limitation of MD simulations ?
No chemistry
No breaking or forming of bonds
No pH
Low complexity systems
Single lipids in membranes
No natural competing interactions
Small systems / simple interactions
Sampling
Local minima
Butterfly effect
Terminology for MD simulations ?
Trajectory
the simulation “movie”
Topology
a description of which atoms are
in the simulation and how they are
connected
Force field
a list of forces for all kinds of atoms.
RMSD
”Root mean square distance”, a
measure of how similar a configuration
of points are to a reference
configuration
Enhanced sampling methods?
Bigger motions and collective changes ( like comformational changes and protein tumbling )
takes longer to observe
“Stuck in local minima”
A high barrier
=
More unlikely
to happen
=
Takes longer
to happen
spontaneously
Transitions between protein states also
has barriers
X-ray structure of receptor
in Apo state
+
A drug
Expect:
A simulation that shows
the transition to the
activated receptor
Pulling simulations
Attach the molecule being pulled to a
spring, which is connected to a “fake” atom
that moves along some interesting motion
motion.
Umbrella sampling
Attach the molecule being pulled to a
spring, which is connected to a “fake”
atom that moves along some interesting
motion motion forces the protein to
sample a particular part of the transition
Analysis methods (MD simulations)?
“Pairwise atom distance” RMSD
NEEDS alignment to be meaningful
Markov State Modeling represents the conformational space of a molecular system as a set of discrete states, where each state corresponds to a distinct region of the system’s configurational space.
Transitions between these states are governed by transition probabilities, which describe the likelihood of transitioning from one state to another in a given time interval.
Principle component analysis
Identification of Dominant Motions: PCA identifies the principal components (PCs) that describe the most significant collective motions in the system. Each PC represents a linear combination of atomic displacements, capturing correlated motions across the molecule.
what are WNTs and what do they bind ?
Signaling Pathway: The WNT signaling pathway is a complex network of intracellular signaling cascades initiated by binding of WNT ligands to cell surface receptors. This pathway regulates diverse cellular processes such as cell proliferation, differentiation, migration, polarity, and stem cell maintenance.
WNT Ligands: The WNT family comprises 19 members in humans, each encoded by a distinct gene. These proteins are characterized by their conserved cysteine-rich domain and are classified into several subgroups based on their sequence homology and functional properties.
Receptors: WNT ligands bind to cell surface receptors belonging to the Frizzled (FZD) family of seven-pass transmembrane proteins. In addition to FZD receptors, WNT signaling can also be modulated by co-receptors such as LRP5/6 (Low-Density Lipoprotein Receptor-related Protein 5/6) and ROR1/2 (Receptor Tyrosine Kinase-like Orphan Receptor 1/2).
WNT/-catenin pathway
- Multiprotein complex
- Signal initiation independent
of conformational dynamics
what is the ternary complex model ?
The ternary complex model is a conceptual framework used to describe the interaction between a ligand (L), a receptor (R), and an effector (E) in pharmacology and signal transduction. This model provides insights into the functional consequences of ligand-receptor interactions and how they may lead to downstream biological effects. The ternary complex model is particularly relevant in the context of G protein-coupled receptors (GPCRs), which are a large family of cell surface receptors involved in various signaling pathways.
Previous work – biochemistry and microscopy on Frizzled ?
Dual color FRAP
* Assesses proteinprotein
interaction
* Crosslinking‐based
FRAP stands for Fluorescence Recovery After Photobleaching. It is a microscopy technique used to study the dynamics of fluorescently labeled molecules within living cells or biological tissues. FRAP provides insights into processes such as protein diffusion, membrane dynamics, and molecular interactions in real-time.
Overhanging aim(s) for the Frizzled receptor ?
Understand WNT-binding
Define the role of the linker domain
Understand parameters
of signal specification
Define mechanisms of effector coupling
Explore the drugability of FZDs
Conclusion - Y2502.39 project, why is it important ?
- Y2502.39 is not phosphorylated but crucial for stabilizing the
receptor structure - Mutation of Y2502.39 affects DVL interaction but not G protein
coupling - First evidence for conformational selection of pathway specifcity
On the hunt for protein
phosphorylation what methods where moslty used?
Signaling
* On the hunt for protein
phosphorylation using MS
and protein chemistry
Bioinformatics:
* Protein modeling
* Molecular dynamics
simulations (distance
measurements)
WNT stimulation leads to
the transient dissociation
of FZD6 dimers. How was that shown ?
Bioinformatics:
* Protein modeling
* Molecular dynamics
simulations
* Protein-protein contact
Conclusion - FZD6 dimer story ?
- WNTs induce a dynamic and reversible dissociation of the FZD6
dimer - The methodology allowed for the first time to assess dimer
dynamics in living cells in the time course of minutes. - The dimer of FZD6 appears to stabilize an inactive conformation
whereas the monomeric form is the conformation that mediates
G protein signaling. - It remains unclear whether this concept is generally applicable
for all FZDs
- Cancer mutations predict
switch mechanism, methods ?
Bioinformatics:
* Sequence alignment
* Cancer database mining
* Protein modeling
* Molecular dynamics
simulations (in silico mutation;
distance measurements)
Conclusion – Molecular Switch
- Class F has a conserved molecular switch mechanism (R6.32-
W7.55) essential for G protein activation - Mutation of the molecular switch (unlocking the interaction)
impairs DVL coupling
Are FZDs druggable with
small molecules?
(binding, mode of action,
efficacy), methods?
Bioinformatics:
* Sequence alignment
* Protein modeling
* Molecular dynamics
simulations
* Volumetric analysis
* Ligand docking
Conclusion – SAG1.3 acting on FZDs ?
- Targting FZDs with small molecules is possible
- SAG1.3 is the first but still a lousy partial agonist at FZD6
- The FZD core contains a ligand binding pocket arguinig again
for intrinsic existence of activation receptor core dynamics
State-stabilizing microswitches, methods used to assess them?
Bioinformatics:
* Sequence alignment
* Protein modeling and
mutagenesis
* Molecular dynamics
simulations
* Volumetric analysis
Conclusion – Molecular switches in Class F ?
- Mutagenesis approach allows definintion of sites defining
effector coupling specificity - FZDs prefer coupling to DVL over G proteins
- Receptor dynamics are essential for pathway selection
- Explanatory model for
FZD activation , methods used?
Bioinformatics:
* CryoEM analysis
* Protein modeling
* Evolutionary analysis
* Molecular dynamics
simulations
* Volumetric analysis
Conclusion – inactive FZD7 CryoEM structure ?
- Molecular switch acts very differently and explains why FZDs
couple to DVL and less well to heterotrimeric G proteins - Water networks can be resolved and they turn out to be
important for receptor activation - Conserved cholesterol binding site is of functional relevance
Disheveled (DVL) proteins ?
Disheveled (DVL) proteins are a family of cytoplasmic proteins that play critical roles in transducing signals from Wnt receptors, particularly Frizzled (FZD) receptors, to downstream signaling pathways. DVL proteins are key mediators of the Wnt signaling pathway, which is essential for various developmental processes, tissue homeostasis, and disease mechanisms.
Methodological considerations
* Bioinformatics guided ex silico
methods ( what can we do with them )?
- Mutagenesis
- Sensor design and sites for
bioorthogonal labeling - Validate dynamics
- Relate receptor function to cancer
- Identify FZD‐targeting compounds
- Allow class F‐wide conclusions
- Integrated use of methodology
- Understanding activation and
signal specification - Buidling new sensors
- Finding FZD‐targeting molecules
