Approaches to approximate molecular states from CryoEM data

David Herreros, PhD Student, Biocomputing Unit (CNB)

CryoEM flexibility estimations rely on two different approaches that can be followed to represent any conformational state represented by a particle, volumen, or structural model. Although both approaches try to solve the same problem, the kind of information they can extract is sliglty different

How can we approximate conformational states?

Density-based approximation

Density-based approximation

• Density based approaches are similar to a standard reconstruction

• Each state is estimated independently, continuity rely on the amount of states that can be approximated

• Therefore, density based approache are advanced reconstruction methods specfically designed to recover many underlying 3D objects

Density-based approximation

Classical 3D classificaction usually done in CryoEM belongs to the density estimation approach

Density-based approximation

Classical 3D classificaction usually done in CryoEM belongs to the density estimation approach

Class 1

Class 2

However this process cannot be continuous… we need much more states

Density-based approximation

Ideally, we have one state per particle, is it posible to do a reconstruction just from a single particle?

?

Density-based approximation

We present below some of the softwares estimating density maps to approximate conformational states from different CryoEM data

CryoDRGN

HetSIREN

GMM

Density-based approximation

Deformation field-based approximation

Deformation field-based approximation

• Deformation fields defined the motion to be abblied to some point to reach the desired target location

• In the proteins case, each point (atomic position or voxel) is moved to its corresponding position in a new conformation

• Therefore, deformation field approaches need a starting point and a target

Deformation direct warping of continuous coordinates (structural models)

Continuous coordinates application

Deformation fields image/volumen warping followed by a proper interpolation scheme

Image/Map application

Deformation field-based approximation

Deformation field approximation is characterized by a large versatiliy:

Deformation direct warping of continuous coordinates

Continuous coordinates application

Deformation fields image/volumen warping followed by a proper interpolation scheme

Image/Map application

Deformation field-based approximation

Deformation field approximation is characterized by a large versatiliy:

Deformation direct warping of continuous coordinates

Continuous coordinates application

Deformation fields image/volumen warping followed by a proper interpolation scheme

Image/Map application

Deformation field-based approximation

Deformation field approximation is characterized by a large versatiliy:

Deformation direct warping of continuous coordinates

Continuous coordinates application

Deformation fields image/volumen warping followed by a proper interpolation scheme

Image/Map application

Deformation field-based approximation

Deformation field approximation is characterized by a large versatiliy:

Deformation field approach allows to work with any CryoEM data type

• CryoEM density maps (structure mapping, rotation/strain analysis…)

• CryoEM particles (conformational landscape extimations)

• Structural models (direct “flexible fitting” of conformational states)

Deformation field-based approximation

We present below some of the softwares estimating deformation fields to approximate conformational states from different CryoEM data

Normal Mode Analysis

Zernike3D

3DFlex

Deformation field-based approximation

Although both approaches try to solve the same problem, the kind of information they can extract is slightly different

Differences between the two approaches

Density based

Deformation field based

Comp/Cont heterogeneity

Small transitions are hard to see

Only maps

Continuous heterogeneity

Small transitions easier to see

Structural models and maps

Less robust to noise

More robust to noise

Any questions?