CryoSPARC Guide
Search
K

Job: Local Refinement (Legacy)

Local refinement (Legacy).

Description

Local Refinement presents a modification to the high-resolution refinement algorithm, that allows particle poses to deviate slightly from their initial alignments. The primary use case for Local Refinement is for refining sub-regions within the consensus volume, which occurs when the complex of interest has suspected motion between sub-units. In these cases, a consensus volume may have clear regions of lower resolution, or may appear "blurred" in certain areas. In addition to this, local refinement can also be used whenever a local alignment search is desired, even using a mask that covers the entire volume. This may be useful for particularly small proteins, and for proteins that are approximately symmetric but have some flexibility or symmetry-breaking features.
In Local Refinement, particle poses are iteratively refined and recomputed by aligning them with the masked sub-volume, leading to improved resolutions in the final structure of the sub-volume. Local refinement may also provide insight into the range of motion that is present in the initial particle dataset, by characterizing the change in pose for each particle, relative to their initial poses.

Input

  • Particles (blob, ctf, alignments3D)
    • Particles are commonly taken as outputs from a Particle Subtraction job, such that only the signal from the desired region is retained in each particle image
    • Since Local Refinement only enables particle poses to deviate slightly from their initial values, it is unique amongst the refinement jobs in that alignment information is required to compute the particles' initial poses
    • Note that particles must have been previously assigned a gold-standard FSC half-set split, which will be the case if the particles are inputted from any of the refinement jobs
  • Volume (map, map_half_A, and map_half_B)
  • Mask

Output

  • Volume (refined sub-volume)
  • Particles
  • Mask (dynamically refined)
  • Plots, including overall orientation distributions, and distributions of the deviances from initial poses

When to use Local Refinement

Local Refinement is ideal for refining a protein with known or suspected flexibility, by explicitly accounting for the motion between the masked sub-volume and the rest of the structure. If an initial map is generated and refined using Homogeneous or Non-Uniform Refinement, and there appears to be a disordered region of the protein in the resulting map, a good first step would be to run 3D Variability and 3D Variability Display to visualize the range of motion that exists in the sample. This may reveal that certain regions of the volume can move relative to each other.
Local Refinement is based upon modelling the sub-volume as a rigid body that rotates around a specified fulcrum point. Thus, rotational motion of a fairly rigid sub-volume will naturally be best resolved using Local Refinement, as opposed to a global bending, twisting, or otherwise flexible motion. Local Refinements are most successful when the sub-volume to be refined contains the largest region of the specimen that is suspected to be in rigid motion relative to the rest of the volume.
In the case study on the Yeast U4/U6.U5 tri-snRNP (spliceosome) dataset, two candidate regions for local refinement of the spliceosome are shown below.
Two candidate regions for local refinement within the snRNP complex.
After arriving at a consensus refinement that has known or suspected relative motion, a typical Local Refinement workflow might look like the following:
  1. 1.
    Using UCSF Chimera, generate masks for the region of the structure you want to refine, and the region that will be subtracted. More information can be found in the Mask Selection and Generation in UCSF Chimera page, which details various methods of generating masks from a consensus refinement in UCSF Chimera, as well as suggested post-processing using the Volume Tools job.
  2. 2.
    (Optionally) Use the Particle Subtraction job to subtract unwanted signal from particles
  3. 3.
    Select a fulcrum (see Fulcrum Selection section below)
  4. 4.
    Run a Local Refinement job using the (subtracted) particles, the consensus refinement volume, and the processed masks, to refine the region of interest
  5. 5.
    Sharpen the results to improve the interpretability of the maps
An example of this workflow on the snRNP dataset is shown in the job tree below.
Example workflow involving Ab-initio Reconstruction and (Non-Uniform) Refinement together with Local Refinement and Particle Subtraction.

Common Parameters

  • Fulcrum Coordinates: the coordinates of the fulcrum must be provided, which is interpreted as the point around which the sub-volume can rotate
    • If left empty, the fulcrum will be located at the center of the box
    • Note: the fulcrum is indexed from the corner of the structure (thus [0,0,0] will correspond to the corner of the structure, rather than the center)
  • Local shift search extent (pix): the translational extent, in pixels, to which particles are allowed to deviate from their initial poses
  • Local rotation search extent (degrees): the rotational extent, in degrees, to which particles are allowed to deviate from their initial poses
  • Alignment Resolution (degrees): the finest alignment search grid spacing to use, which will be enforced by the final refinement iteration
  • Use Non-Uniform Refinement: Optionally, Non-Uniform Refinement can be enabled to help resolve molecules with spatially-varying local resolution. This is particularly useful for membrane proteins, or any proteins with disordered regions (e.g. micelles)
  • Mask (dynamic, static): This can be set to dynamic to re-generate a dynamic mask at each iteration of the refinement, or instead to static to use the unmodified input mask at each iteration.
    • Dynamic masking parameters (if dynamic masking is enabled):
      • Dynamic mask threshold (0-1): The level-set threshold for selecting regions to include in the dynamic mask
      • Dynamic mask near (A): The distance to dilate the dynamic mask to, after thresholding
      • Dynamic mask far (A): The distance at which the mask becomes 0.0
      • Note that the width of the soft padding is equal to the difference between the far and near distances. For smaller masks, it is often desirable to make the masks softer (by increasing the padding width) in order to limit overfitting
For tips on the tuning of these parameters, please see the Fulcrum Selection and Search Extents sections below.

Fulcrum Selection

Generally, the results of a Local Refinement are sensitive to the fulcrum location, as it defines the center of the alignment search space over poses and translations. If refining a sub-volume, the fulcrum could be placed within the boundary between the sub-volume and the rest of the volume. Alternatively, many users have found success by placing the fulcrum at the centroid of the mask, which can be computed in UCSF Chimera via the measure center command, which reports the center of mass of the mask in grid-indices. Below shows an example of using Chimera to find the center of mass of a mask (in this case, the mask covers the "head" of the tri-snRNP complex).
Example of measuring the center of mass of a mask in UCSF Chimera
If you would like to find coordinates of an arbitrary point on the volume in UCSF Chimera, for example on the boundary of a mask, the Volume Tracer tool can be used in Tools > Volume Tracer. To find the coordinates of a particular point on the boundary of a mask, In the Volume Tracer panel, under "Mouse", select "Place markers on surfaces". Then, using the mouse button indicated, click on the region of the mask that you want to get the coordinates of. Finally, to get its coordinates, type getcrd sel in the command line. Note that unlike the measure command, coordinate values are given in Angstroms, so these values must be divided by the pixel size in order to get the fulcrum coordinates in voxels (which is what Local Refinement expects). Below shows an example of placing the fulcrum on the boundary between the tri-snRNP "body" and "head" region (not shown).
The fulcrum point is marked by the small green sphere on the mask surface, and its coordinates are shown at the bottom, in the reply log.
If refinements with the fulcrum located at the mask center of mass (or on the mask boundary) don't seem to produce ideal results, it may help to increase the local shift and rotation search extents in order to capture a larger alignment search space. Note that internally, the volume is stored by shifting it to be centered at the fulcrum's coordinates, so fulcra that do not lie within the original volume (or fulcra that lie near the edge of the box) should be avoided.
In any case, Local Refinement is sensitive to the fulcrum coordinates, as well as to the mask chosen. Multiple rounds of local refinement with varying fulcrum coordinates, as well as varying local search extents, may affect the quality of results.

Search Extents

Generally, Local Refinements are also sensitive to the search extents over rotations and translations. As a first step, it is best to use prior knowledge about the extent of motion present in the dataset (for example, using 3D Variability and 3D Variability Display jobs), but often multiple refinements with varying search extents may reveal the best choice.
At each iteration of the refinement, histograms in the stream log show the distribution of particle images over the magnitude of the rotation/translational change. Early on in the refinement, when the structure is still at a relatively low resolution, these histograms may show strong maxima at extremal values of the search ranges, as shown below.
Peaked distributions in shifts early on in the refinement
This is acceptable early on in a refinement. As the refinement progresses, these strong peaks should ideally smoothen out. If by the final iteration, there are still strong peaks in the shift or rotation distributions, this may point towards either:
  • a suboptimal choice of search extents, and/or
  • a suboptimal fulcrum selection
In this case, further refinements with better fulcra locations and/or expanded search extents may increase the final refinement resolution.
An ideal local rotation/shift distribution, at the final iteration of a local refinement

Common Next Steps

  • Download and inspect map

Case Study

For a detailed guide to local refinement in CryoSPARC, see the Case Study: Yeast U4/U6.U5 tri-snRNP.