Structure prediction and crystal

X-ray diffraction

Juan J. Martínez Irujo

Departamento de Bioquímica y Genética
Asignatura: “Estructura de Macromoléculas”
Curso: 4º de Bioquímica. 2023 - 2024

Protein Structure Determination

 Introduction: why bother with structure?
 Protein Structure Determination
1. Computational prediction
1. Template Modeling
• Homology Modeling
• Threading
2. Template-Free Modeling (ab initio Methods)
• Physics-Based
• Knowledge-Based

2. Experimental Methods
2.1. X-ray diffraction and crystallography
2. Nuclear Magnetic Resonance
3. Cryogenic Electron Microscopy

Index
Protein Structure is Hierarchical

Introduction: why bother with

structure?

Structure Determines Function

Introduction: why bother with

structure?
 Large scale assemblies

• Multimeric Proteins
• Macromolecular Assemblies
Ribosome:
Protein
Synthesis

Hemoglobi
n: A Replisome:
tetramer DNA
copying
Introduction: why bother with
structure?

Protein Structure Determination

Introduction: why bother with

structure?
Experimental Protein Structure
Determination
1 X-ray crystallography
• Most accurate
• Requires crystals (“in vitro”)
• Able to trace amino acid side chains
• Time consuming and expensive
• Used to determine 86% of structures

2 NMR
• Magnetic field applied to proteins in solution (“in
vivo”)
• Limited to small proteins: 350 amino acids (40
kDa)
• Does not require crystallization
• Time consuming and hardly

3 Electron Microscopy
• Imaging technology
• Low resolution

Introduction: why bother with

structure?

Introduction: why bother with

structure?
 Datos a 30 de enero de 2024

84,6% 8,7% 6,6%

Introduction: why bother with

structure?

Introduction: why bother with

structure?
 Protein Structure Determination

 Introduction: why bother with structure?

 Protein Structure Determination
1. Computational prediction
1. Template Modeling
• Homology Modeling
• Threading
2. Template-Free Modeling (ab initio
Methods)
• Physics-Based
• Knowledge-Based

2. Experimental Methods

Inde
x

Protein structure: computational

prediction
Major Techniques:

1. Template Modeling
• Homology Modeling
• Threading
• Both use known protein structure

2. Template-Free Modeling (ab initio Methods)

• Physics-Based
• Knowledge-Based
• without use of known protein structure

1. Computational
prediction
 Homology modelling

Homology Modelling
• also called comparitive modeling.
• predict protein structures based on sequence homology with known
structure.

Principle:
• if two proteins share a high enough sequence similarity, they are likely
to have very similar three dimensional structure.
• modeling server:-modbase,swiss-model etc.
• Fail in absence of homology (<30%)

1. Computational prediction: Template

modeling

Homology modeling

 Homology modeling is, to date, the most reliable and well-

established
computational approach for predicting protein structures.

 In this method, one or more experimental three-dimensional

structures of related homologous proteins are identified and
used as templates, based on which an atomic-resolution
model of the "target" protein is built from its amino acid
sequence.

 It has been shown that three-dimensional protein structure is

evolutionarily more conserved than would be expected on the
basis of sequence conservation alone.

 Thus, even proteins that have diverged appreciably in

sequence but still share detectable similarity will also share
common structural properties, particularly the overall folding.
1. Computational prediction: Template
modeling
 Homology modeling

o Find template
o Align target sequence with
template
o Generate model:
• add loops
• add side-chains
o Refine model

1. Computational prediction: Template

modeling

Protein Threading (fold recognition)

Places the residues of unknown 'P' along the backbone of a

known structure and determines stability of side chains in
that arrangement

1. Computational prediction: protein

threading
 Protein Threading (fold recognition)

• Protein threading predicts the structural fold by fitting the sequence

into a structural database (PDB) and selecting the best fitting fold.

• Is useful method for predicting the structure especially when the

protein shares a low sequence level identity (i.e. <25%) with other
proteins. We can identify structurally similar proteins even without
detectable sequence similarity.

• The PDB statistics shows that in recent years, only a limited

number of completely new protein folds appear although several
thousand new structures are deposited to the PDB

• These observations imply that in principle, a reasonably good

template can be identified for most proteins with unknown
structures if a perfect protein threading protocol is available.

1. Computational prediction: protein

threading

Protein structure: computational

prediction
Major Techniques:

Template Modeling
• Homology Modeling
• Threading
• Both use known protein structure

Template-Free Modeling (ab initio Methods)

• Physics-Based
• Knowledge-Based
• without use of known protein structure

1. Computational prediction: Template-free

modeling
Computational prediction: overall
approach

1. Computational prediction: Template-free

modeling

Ab Initio Protein Structure Prediction

1. Computational prediction: Template-free

modeling
Ab Initio Protein Structure Prediction

1. Computational prediction: Template-free

modeling

Protein Structure Determination

1. Computational prediction: Template-free

modeling
 Ab Initio Protein Structure Prediction
• Ab initio protein structure prediction methods build protein 3D
structures from sequence based on physical principles.
• Importance
• The ab initio methods are important even though they are
computationally demanding
• Ab initio methods predict protein structure based on physical models,
they are indispensable complementary methods to Knowledge-based
approach.
• Knowledge-based approach would fail in following conditions:
• Structure homologues are not available
• Possible undiscovered new fold exists.

Anfinsen’s theory: Protein native structure corresponds to the state with the lowest
free energy of the protein-solvent system.

1. Computational prediction: Template-free

modeling

Ab Initio Protein Structure Prediction

1. Computational prediction: Template-free

modeling
 The protein folding funnel

1. Computational prediction: Template-free

modeling

• Through an enormous experimental effort, the structures of around 100,000 unique

proteins have been determined, but this represents a small fraction of the billions of
known protein sequences…
• Predicting the three-dimensional structure that a protein will adopt based solely on its
amino acid sequence—the structure prediction component of the ‘protein folding
problem’—has been an important open research problem for more than 50 years.
• Despite recent progress, existing methods fall far short of atomic accuracy, especially
when no homologous structure is available.
• Here we provide the first computational method that can regularly predict protein
structures with atomic accuracy even in cases in which no similar structure is known.

1. Computational methods: Template-free

modeling
 (the prediction was made using AlphaFold without
intervention)

1. Computational prediction: Template-free

modeling

Google Deepmind: AlphaFold

2000 Start-up fundada en 2010, por Demis Hassabis

La compañía creó una red neuronal que aprende a jugar videojuegos de
manera similar a los humanos

2014 Google compró Tecnologías DeepMind creando Google DeepMind

2015 Un programa llamado AlphaGo, impulsado por DeepMind, venció al

campeón europeo de Go. Esta es la primera vez que una inteligencia
artificial (AI) venció a un jugador profesional de Go.

2017 El equipo de DeepMind logró un gran avance al derrotar el módulo

AlphaZero a Stockfish 8 en una serie de cien partidas. Para lograrlo, tan
solo necesitó conocer las reglas del juego y 4 horas de entrenamiento
jugando contra sí mismo. En 2017, StockFish 8 era el campeón vigente
del campeonato mundial de ajedrez por computadora, con un ELO de
3400 puntos.

1. Computational prediction: Template-free

modeling
 Google Deepmind: AlphaFold

Aprendizaje reforzado profundo

• A diferencia de otras IA, como Deep Blue o Watson de IBM, las cuales fueron
desarrolladas para un propósito predefinido y solo funcionan dentro de su
ámbito, Google DeepMind dice que sus sistemas no están pre-programados:
aprenden de la experiencia.

• Prueban el sistema en videojuegos, en especial en juegos arcade, como

Space Invaders o Breakout. Sin alterar el código, la IA empieza a entender
como jugar el juego, y después de jugar varias veces, en algunos juegos
(especialmente en Breakout), tiene una mayor eficiencia al jugar que
cualquier humano.

• El objetivo de Tecnologías DeepMind es "resolver la inteligencia", lo que

están tratando de lograr mediante la combinación de "Las mejores técnicas
de Aprendizaje automático y Neurociencia de sistemas para construir
potentes algoritmos de aprendizaje de propósito general"

1. Computational prediction: Template-free

modeling

2018. AlphaFold 1, obtuvo el primer lugar en la clasificación general de la 13ª

edición de la competición CASP (Critical Assessment of Techniques for
Protein Structure Prediction, 'evaluación crítica de técnicas para la
predicción de la estructura de proteínas’)

2020 AlphaFold 2 volvió a ganar la competición CASP en noviembre de 2020

En alrededor de dos tercios de las proteínas, AlphaFold 2 obtuvo una puntuación

superior a 90; una puntuación de 100 denota una coincidencia
completa.

Los resultados de AlphaFold 2 en CASP han sido calificados como

«asombrosos» y «transformadores». Aunque la exactitud de las
predicciones no es lo suficientemente alta en un tercio de los casos y el
programa no revela ninguna información sobre el mecanismo del
plegamiento de proteínas, el logro técnico ha recibido un reconocimiento
generalizado.

1. Computational prediction: Template-free

modeling
 Google Deepmind: AlphaFold

[Link]

1. Computational prediction: Template-free

modeling

Protein Structure Determination

 Introduction: why bother with

structure?
 Protein Structure Determination
1. Computational prediction

2. Experimental Methods
1. X-ray diffraction and
crystallography
2. Nuclear Magnetic Resonance
3. Cryogenic Electron Microscopy

Inde
x
Protein Structure Determination

A. Introduction: why bother with

structure?

Experimental Methods

1. X-ray diffraction and

crystallography
• Overview
• What is diffraction? Why do we get
spots?
• Fiber vs crystal diffraction
• X-ray instrumentation
• The phase problem in X-ray
crystallography
• Electronic density and refinement
• Structure validation
•

2.1 X-ray diffraction.

Overview of the X-ray Crystallographic
Method

2.1 X-ray diffraction.

Overview

Optical Microscope vs X-Ray

Cristallography

2.1 X-ray diffraction.

 Optical Microscope vs X-Ray
Cristallography

2.1 X-ray diffraction.

Overview

Optical Microscope vs X-Ray

Cristallography

Comparación entre un microscopio óptico y la difracción de los rayos X en

términos de lo que podría denominarse el "microscopio imposible" de rayos X, ya
que no existen lentes capaces de recombinar los rayos X.

2.1 X-ray diffraction.

 Brief history of X-ray crystallography
1895 Wilhelm Röntgen, a German physicist, discovered X-rays and
demonstrated that X-rays could pass directly through human
tissues leaving behind images of bones. Nobel Prize in Physics
in 1901.
1912 Max von Laue, another German physicist, proved that crystals
could diffract X-rays. He won the Nobel Prize in Physics in
1914.
1913 William Henry Bragg and William Lawrence Bragg (father and son)
solved first X-ray structure, using table salt. They shared the
1915 Nobel Prize for their pioneering efforts.
1926 First enzymes crystalized. Sumner crystalized urease (1926) and
Northrop crystalized pepsine (1930). Received the 1946 Nobel
Prize in Chemistry.
1959 Structure of the first protein myoglobin by Perutz and
Kendrew.
Nobel Prize in 1962 in Chemistry.
1962 Francis Crick, John Watson, and Maurice Wilkins received the
Nobel Prize by the structure of DNA (19539, based on X-ray
diffraction patterns of the DNA double helix produced by
Rosalind Franklin.
1965 Structure of the first enzyme: lysozyme
Wilhem Röntgen
2.1 X-ray diffraction. Overview

2.1 X-ray diffraction.

 Max Laue: diffraction of inorganic
cystals

NaCl(s)
Crystal structure
with sodium in
purple and chloride
in green

2.1 X-ray diffraction.

Overview

Kendrew: diffraction of sperm whale

myoglobin

Kendrew's original model of

the myoglobin molecule

2.1 X-ray diffraction.

 Experimental Methods

1. X-ray diffraction and

crystallography
• Overview
• What is diffraction? Why do we get
spots?
• Fiber vs crystal diffraction
• X-ray instrumentation
• The phase problem in X-ray
crystallography
• Electronic density and refinement
• Structure validation
•

2.1 What is diffraction? Why do we get

spots?

Light Reflection – Refraction -

Diffraction
Reflection is the change in direction of a
wavefront when it hits an object and
returns at an angle

Refraction occurs when the wavefront changes

direction as it passes through a medium. The
degree of refraction is dependent upon the
wavelength of light and the index of refraction
of the medium.

Diffraction usually occurs when waves

pass through a narrow slit or around a
sharp edge.

2.1 What is diffraction? Why do we get

 Light Diffraction

The scattering process bends light’s direction of

propagation. The degree of bend is dependent on the
relative difference between its wavelength and size of the slit.

•If the slit size is smaller than or equal to light’s

wavelength, considerable diffraction occurs. Smaller the
slit, greater the degree of diffraction.

•If the slit size is larger than light’s wavelength, diffraction

still occurs but is unnoticeable to the naked eye.

2.1 What is diffraction? Why do we get

spots?

Young's Double-slit Experiment

(1801)

2.1 What is diffraction? Why do we get

 Wave interference

2.1 What is diffraction? Why do we get

spots?

Double-slit Experiment

2.1 What is diffraction? Why do we get

 Double-slit Experiment

El mismo conjunto de doble rendija (0,7 mm entre las rendijas); en la imagen superior, una
rendija está cerrada. En la imagen de una sola rendija, se forma un patrón de difracción
(los puntos débiles a ambos lados de la banda principal) debido al ancho distinto de cero
de la rendija. También se ve un patrón de difracción en la imagen de doble rendija, pero
con el doble de intensidad y con la adición de muchas franjas de interferencia más
pequeñas.

2.1 What is diffraction? Why do we

get spots?

Max Laue: diffraction of inorganic

cystals

NaCl(s)
Crystal structure
with sodium in
purple and chloride
in green

2.1 What is diffraction? Why do we get

Difraction from oriented fibers of
DNA

2.1 What is diffraction? Why do we get

spots?

Constructive and destructive

interference

2.1 What is diffraction? Why do we get

 Why X-rays?

Optics: Limit of resolution ~ /2

Visible light (400 -700 nm) 0,4 – 0,7

µm
- Optical microscopy resolution : 0,2 µm

Distance between atoms: 0.1 – 0.3 nm (1 -

3 Å)
Wavelenght of X-rays: 0.1 – 10 nm (1 - 10
Å)

X-ray generators produce photons of λ = 0,5

– 2,5 Å
 Allows structure determination at
atomic resolution

Problem:
• No lenses available for X-rays (as for visible
light in microscope)
X-rays
• Unlike visible light, X-rays cannot be focused by lenses. The
refractive index of X-rays in all materials is very close to 1.0.
• X-rays are higher energy wavelengths of EMR used in
2.1 What is diffraction? Why do we get
crystallography, medical scanning, and other applications
spots?
(airport security scanning).

2.1 What is diffraction? Why do we get

 Why can’t we visualize molecules
• directly?
A single molecule is a very weak scatterer of X-rays. Most of the X-
rays will pass through the molecule without being diffracted. The
diffracted rays are too weak to be detected.

• Solution: Analyzing diffraction from crystals instead of single

molecules. A crystal is made of a three-dimensional repeat of
ordered molecules (1014) whose signals reinforce each other. The
resulting diffracted rays are strong enough to be detected.

2.1 What is diffraction? Why do we get

spots?

Bragg’s Law
Why do we get spots (reflections) and not a
diffuse pattern of scattered x-rays?

distance between successive

parallel planes (the
"interplanar" spacing)

Difference in path must equal integral number of wavelenths

(nλ)
• This is the condition for successful reinforcement of
waves reflected of different 2.1 What is diffraction? Why do we get
layers
Bragg’s Law

• Derived by English physicists

Sir William Henry Bragg and
his son Sir William Lawrence
Bragg

• Shared Nobel Prize in Physics,

1915

2.1 What is diffraction? Why do we get

spots?

Goniometer

2.1 What is diffraction? Why do we get

 X-ray Protein Crystallography

2.1 What is diffraction? Why do we get

spots?

X-ray Protein Crystallography

1. Make crystals of your protein

 0.3 – 0.1 mm in size
 Proteins must be in an ordered, repeating pattern

2. X-beam is aimed at crystal and data is collected

3. Structure is determined from the diffraction data
2.1 What is diffraction? Why do we get
 X-ray Protein Crystallography

2.1 What is diffraction? Why do we get

spots?

Experimental Methods

1. X-ray diffraction and

crystallography
• Overview
• What is diffraction? Why do we get
spots?
• Fiber vs crystal diffraction
• X-ray instrumentation
• The phase problem in X-ray
crystallography
• Electronic density and refinement
• Structure validation
•

2.1 Fiber vs crystal

 X-ray fiber diffraction

Fibrous materials such as wool or cotton easily form aligned bundles,

and were among the first biological macromolecules studied by X-ray
diffraction, notably by William Astbury in the early 1930s. Fiber
diffraction data led to several important advances in the
development of structural biology, e.g., the original models of the α-
helix (α-queratina), the collagen triple helix and the Watson- Crick model
of double-stranded DNA.

2.1 Fiber vs crystal diffraction

Diffraction from oriented fibrous

proteins Fibrous molecules rarely crystallize and
hence are usually not subject to
structural determination by single-crystal
X-ray structure analysis (Section 8-3A).
Rather than crystallizing, they associate as
fibers in which their long molecular axes are
more or less parallel to the fiber axis but in
which they lack specific orientation in
other directions. The X-ray diffraction
pattern of such a fiber, example, contains
little information, far less than would be
obtained if the fibrous protein could be
made to crystallize. Consequently, the
structures of fibrous proteins are not
known in great detail. Nevertheless, the
original X-ray studies of proteins were
carried out in the early 1930s by William
Astbury on such easily available protein
fibers as wool and tendon. Since the first X-
ray crystal structure of a protein was not
determined until the late 1950s, these fiber
studies constituted the first tentative steps
in the elucidation of the structural principles
governing proteins and formed much of
the experimental basis for Pauling’s
formulation of the helix and pleated
sheet. 2.1 Fiber vs crystal
Diffraction from oriented fibers of DNA

2.1 Fiber vs crystal

diffraction

Diffraction from oriented fibers of DNA

2.1 Fiber vs crystal

 Single Crystal X Ray Diffraction Studies of
DNA
Oligonucleotides

Normalmente, cuando se piensa o

habla del ADN, uno se lo imagina
como una doble hélice perfecta,
completamente regular. En realidad,
la estructura de ADN presenta
variaciones y desviaciones locales.

2.1 Fiber vs crystal

diffraction

Experimental Methods

1. X-ray diffraction and

crystallography
• Overview
• What is diffraction? Why do we get
spots?
• Fiber vs crystal diffraction
• X-ray instrumentation
• The phase problem in X-ray
crystallography
• Electronic density and refinement
• Structure validation
•

2.1 X-ray
Protein X-ray crystallography

2.1 X-ray
instrumentation

X-ray instrumentation

2.1 X-ray
X-ray instrumentation

2.1 X-ray
instrumentation

X-ray optics

2.1 X-ray
 X-ray instrumentation

Figure 1 The mechanism of characteristic X-ray generation.

2.1 X-ray instrumentation

X-ray source

2.1 X-ray
 Synchrotron radiation

2.1 X-ray
instrumentation

Properties of Synchrotron Radiation

2.1 X-ray
X-ray instrumentation

2.1 X-ray
instrumentation

Goniometer

2.1 X-ray
 Experimental Methods

1. X-ray diffraction and

crystallography
• Overview
• What is diffraction? Why do we get
spots?
• Fiber vs crystal diffraction
• X-ray instrumentation
• The phase problem in X-ray
crystallography
• Electronic density and refinement
• Structure validation
•

2.1 The phase

problem

X-ray diffraction
• In a single-crystal X-ray diffraction measurement, a crystal is
mounted, then positioned at multiple selected orientations. The
crystal is bombarded with a finely focused monochromatic beam
of X-rays, producing a diffraction pattern of regularly spaced spots
known as reflections.
• Two-dimensional images taken at different rotations are converted into
a three-dimensional electron density model using Fourier transforms,
combined with chemical data known for the sample.
• Poor resolution (fuzziness) or even errors may result if the
crystals are too small, or not uniform enough in their internal
makeup.

2.1 The phase

 Optical Microscope vs X-Ray
Cristallography

Comparación entre un microscopio óptico y la difracción de los rayos X en

términos de lo que podría denominarse el "microscopio imposible" de rayos X, ya
que no existen lentes capaces de recombinar los rayos X.

2.1 The phase problem

El problema de la fase

Intensidad

Transformada de Fourier
2.1 The phase
 El problema de la fase

• Entre los espacios directo y recíproco existe una relación que

matemáticamente es una transformada de Fourier que no podemos
resolver en sentido izquierdo del esquema ya que no conocemos
fases relativas (Φ) de los haces de difracción

• El experimento de difracción no nos proporciona una de las

magnitudes fundamentales para resolver la ecuación, las fases
relativas (Φ) de los haces de difracción

2.1 X-ray
instrumentation

El problema de la fase

Espacio real o directo Espacio recíproco

Intensidad Fase
Transformada de Fourier
2.1 The phase
 Métodos de resolución para
macromoléculas
En cristales que contienen moléculas grandes, proteínas o enzimas,
el problema de la fase puede resolverse mediante tres métodos,
dependiendo del caso:

(i) introduciendo átomos altamente dispersores, o método del

Reemplazo Isomorfo Múltiple (MIR, del inglés, Multiple
Isomorphous Replacement), y Basado en el método de Patterson
(simplificación matemática). (Kendrew – Mioglobina)

(ii) introduciendo átomos dispersores anómalos, o método de

Difracción Anómala Múltiple (MAD, del inglés Multi-wavelength
Anomalous Diffraction) (sin metales pesados, i.e., obtener proteínas
recombinantes en las que los residuos de metionina están
reemplazados por selenio-metionina)

(iii) Mediante el Reemplazo Molecular (MR, del inglés

Molecular Replacement), haciendo uso de un modelo
estructural de una proteína homóloga, previamente
determinada.

2.1 The phase

problem

Isomorphous Replacement (MIR)

2.1 The phase

Método del reemplazo isomorfo múltiple
(MIR)

• El método MIR consiste en introducir, en la celdilla

cristalina, átomos que sean grandes dispersores de los rayos
X (átomos pesados, como Au, Pt, Fe, Ag, Hg).

• The coordinates of the heavy atom can be calculated with

small- molecule-methods. These include so-called direct
methods and Patterson methods.

• Los átomos introducidos no deben distorsionar la estructura

cristalina de la proteína nativa, es decir, que los cristales así
obtenidos deben ser isomorfos de los de la proteína nativa.

• En la práctica, este reemplazo isomorfo se lleva a cabo

difundiendo complejos de metales "pesados" a través de los
canales que poseen los cristales de proteína, y en donde
normalmente existen cadenas laterales de aminoácidos con
capacidad de coordinar a los átomos metálicos (por ejemplo
grupos SH de las cisteínas).

2.1 The phase problem

X-ray instrumentation

2.1 The phase

Soluciones al problema de la fase

2.1 The phase

problem

Experimental Methods

1. X-ray diffraction and

crystallography
• Overview
• What is diffraction? Why do we get
spots?
• Fiber vs crystal diffraction
• X-ray instrumentation
• The phase problem in X-ray
crystallography
• Electronic density and refinement
• Structure validation
•

2.1 Electronic density and

 Resolution of molecular and crystalline
structures using X-ray
diffraction

2.1 Electronic density and

refinement

Resolution and Model Building

2.7Å is a good resolution for a structure solved by X-ray crystallography, but many structures now achieve much
higher resolution: most structures in the PDB are between 1.8Å and 2Å resolution. With cryo-electron
microscopy, it is difficult to achieve such high resolutions and 3.5Å is considered good, as it allows the
visualisation of side chains.
2.1 Electronic density and
 Electron Density Maps – Amino Acid
Residues

Amino acid residues can produce distinctive electron density signatures

at higher resolutions (less than 2.0 Å)

2.1 Electronic
density and
refinement

Refinement

2.1 Electronic density and

 Experimental Methods

1. X-ray diffraction and

crystallography
• Overview
• What is diffraction? Why do we get
spots?
• Fiber vs crystal diffraction
• X-ray instrumentation
• The phase problem in X-ray
crystallography
• Electronic density and refinement
• Structure validation
•

2.1 Structure
validation

Structure Validation

2.1 Structure
 X-ray crystallography today

• It took Kendrew and Perutz almost 20 years to solve the globin

structures (Left).
• Today, with a very good crystal, a protein structure can be solved
in a day. Modelling has become much more sophisticated (Right).

2.1 X-ray
crystallography

Crystallography Pros/Cons

Advantages
• can be “fast” – down to a few months
• Large structures posible (ribosome)
• May give very good resolution (down to
0,5 Å)

Disadvantages
• Requires cristal formation
• Non-physiological conditions
• Crystal contacts can limit protein motion

2.1 X-ray
Brief History of X-ray crystallography

2.1 X-ray
crystallography