complex. However, if individual components cannot be placed into
the map with any confidence, a more thorough search is needed to
rigidly place individual subunits.
One program that can be employed for both local and global
rigid fitting is Mod-EM [25] implemented in MODELLER
[2]. Briefly, given a density map and candidate model, the model
is converted into a candidate density by the blurring procedure
described in Subheading 2.2 above. The best fit is obtained by
altering the position of the map to maximize the CCC between
candidate and cryo-EM density.
When the cryo-EM map corresponds one-to-one with the
density from the candidate model, a local rigid fitting protocol is
performed (similar to the fit-in-map approach in UCSF-Chimera
[3]). This can be the case when the map represents the candidate
model only, or when a protein complex map has been segmented
around an individual protein (e.g., using tools such as Segger [26]
implemented in Chimera [3]). In this case, the candidate model
center-of-mass is translated to the map center-of-mass, and a local
search is performed either exhaustively, or with a Monte Carlo
method. A single step consists of a translation of the probe for a
single grid unit in the map and a search of three Euler angles.
When segmentation of the density corresponding to the candidate model from the map is not possible, a global rigid fit protocol
is employed. There are two available protocols for this in Mod-EM
and their use is dependent on the relative size of the candidate
model to the map. Typically, if the candidate model represents a
significant portion of the map (>35% volume) a Monte Carlo
method is recommended. This begins with an initial arbitrary
placement of the model into the map before executing a number
of Monte Carlo steps. When the candidate model represents a small
volume of the map ( 35% volume), a scanning Monte Carlo protocol is used, where the map is divided into “cells” approximately
the volume of the candidate model and a local search is performed
in each cell for which the probe initial CCC is positive.
Global rigid fitting methods such as that implemented in
Mod-EM generally work best at resolutions where the boundaries
of proteins and domains can be identified in the map. However, at
lower resolutions (>15 A ˚ ) this is not always possible and the use of
CCC can give misleading results regarding the fit of individual
components in the map as their densities overlap upon complex
formation, and without taking this into account global optimization has been shown to worsen the fit [27]. In fact, with advances in
cryo-EM and cryo-ET, it is now realistic to obtain maps for very
large complexes, like the nuclear pore complex [28]. One way
around this problem is to simultaneously optimize the fit of all
the components in the map (namely, assembly/simultaneous/
multi-component fitting).
CryoEM Density Fitting and Validation
195
the map with any confidence, a more thorough search is needed to
rigidly place individual subunits.
One program that can be employed for both local and global
rigid fitting is Mod-EM [25] implemented in MODELLER
[2]. Briefly, given a density map and candidate model, the model
is converted into a candidate density by the blurring procedure
described in Subheading 2.2 above. The best fit is obtained by
altering the position of the map to maximize the CCC between
candidate and cryo-EM density.
When the cryo-EM map corresponds one-to-one with the
density from the candidate model, a local rigid fitting protocol is
performed (similar to the fit-in-map approach in UCSF-Chimera
[3]). This can be the case when the map represents the candidate
model only, or when a protein complex map has been segmented
around an individual protein (e.g., using tools such as Segger [26]
implemented in Chimera [3]). In this case, the candidate model
center-of-mass is translated to the map center-of-mass, and a local
search is performed either exhaustively, or with a Monte Carlo
method. A single step consists of a translation of the probe for a
single grid unit in the map and a search of three Euler angles.
When segmentation of the density corresponding to the candidate model from the map is not possible, a global rigid fit protocol
is employed. There are two available protocols for this in Mod-EM
and their use is dependent on the relative size of the candidate
model to the map. Typically, if the candidate model represents a
significant portion of the map (>35% volume) a Monte Carlo
method is recommended. This begins with an initial arbitrary
placement of the model into the map before executing a number
of Monte Carlo steps. When the candidate model represents a small
volume of the map ( 35% volume), a scanning Monte Carlo protocol is used, where the map is divided into “cells” approximately
the volume of the candidate model and a local search is performed
in each cell for which the probe initial CCC is positive.
Global rigid fitting methods such as that implemented in
Mod-EM generally work best at resolutions where the boundaries
of proteins and domains can be identified in the map. However, at
lower resolutions (>15 A ˚ ) this is not always possible and the use of
CCC can give misleading results regarding the fit of individual
components in the map as their densities overlap upon complex
formation, and without taking this into account global optimization has been shown to worsen the fit [27]. In fact, with advances in
cryo-EM and cryo-ET, it is now realistic to obtain maps for very
large complexes, like the nuclear pore complex [28]. One way
around this problem is to simultaneously optimize the fit of all
the components in the map (namely, assembly/simultaneous/
multi-component fitting).
CryoEM Density Fitting and Validation
195