Molecular Replacement
Why
One of the biggest challenges in crystallography is the phase problem,
which can be addressed with experimental phasing methods. A popular alternative
approach is molecular replacement (MR), which uses known structures to
solve the phase problem. When a structurally similar model (e.g., a homologue)
is available, this model is computationally placed in the unit cell
containing the unknown structure. An initial electron density map is
then calculated with the phases from the homologue and the observed
structure factors.
The success of MR calculations depends on how much signal can be
extracted from the data using the particular model, which in turn
depends on a combination of the model quality and completeness,
data resolution, and number of diffraction observations. For crystals
of medium-sized proteins diffracting to moderate resolution, for
example, the sequence identity between the molecule and homologue
should be >25-30% and the r.m.s. deviation between C⍺ atoms should
be <2.0 Å.
How
In Phenix, the primary program for performing molecular replacement is
Phaser. This requires experimental diffraction data, atomic model(s),
the sequence of the molecule in the crystal, and typically some estimate
of the number of molecules in the crystal. Phaser uses maximum-likelihood
algorithms to determine the rotation of the model(s) with respect to the
unit cell, and then the translation of the rotated model(s) within the unit
cell. The output of this process, if successful, is a model file containing
the placed model(s) and an MTZ file containing coefficients for the electron
density made using the placed model(s) and the observed experimental amplitudes.
For an introduction to molecular replacement in Phenix, click
here.
How to use the Phaser GUI in Phenix: Click
here
Phenix reference manual for Phaser.
Common issues
- No solutions were found even with a good search model: There are multiple
causes, but a common one is that a multiple domain protein has undergone
a conformational change. In this case, you should split the structure and
perform molecular replacement on the individual domains using separate
search models. An alternative cause, thankfully less common, is that the
crystallized molecule is not the intended one. For this reason, it is always
prudent to use the "Search PDB Symmetry" tool in the Phenix GUI to check
the cell dimensions and space group against the PDB. A third possibility is
that the space group is incorrect (e.g., if twinning increases the apparent
symmetry of the diffraction pattern), thus it is always good to run Xtriage
in the "Data analysis" section of the Phenix GUI.
- Frequently asked questions about molecular replacement
Related programs
- phenix.MRage: This program uses Phaser to
perform highly automated molecular replacement. In its simplest form,
it uses experimental diffraction data and a sequence to solve a
molecular replacement problem.
- phenix.sculptor: This program can improve
models for molecular replacement by trimming off parts of the model that
are unlikely to be preserved in the target structure.
- phenix.ensembler: This program prepares
ensembles for molecular replacement. An ensemble is a set of structurally
related models that have been superimposed with respect to each other,
typically using a conserved structural core.
Phenix.ensembler automates the process of superimposing models.
- phenix.morph_model: This program
can improve an initial model after molecular replacement by locally moving
the structure to better fit the electron density map. This is especially
powerful when the molecular replacement solution is structurally too different
to provide phases for map interpretation or automated model building.
- phenix.mr_rosetta: This program can
improve an initial model after molecular replacement by using the Rosetta
program to modify the model such that it moves closer to the true structure,
while also improving the fit to electron density map. This is especially
powerful when the molecular replacement solution is structurally too
different to provide phases for map interpretation or automated model
building.