Job: Particle Subtraction (BETA)


Particle Subtraction performs signal subtraction of unwanted signal in particle images, which is commonly used on input particles prior to a Local Refinement job. The result of a Particle Subtraction job is a particle stack in which the undesired signal (associated with the masked input volume) is absent. In order to do this, the job applies the mask to the volume, isolating the portion of the density map containing the unwanted signal. Then, using the particles' 3D alignment information, the signal from the masked volume is projected, convolved with the contrast transfer function, and subtracted from each particle image.

Inputs, Parameters, and Outputs


  • Particles (blob , ctf, and alignments3D)

    • Note: Since alignment information is required, the particles must have came from a previous (consensus) refinement job

  • Volume (map , or map_half_A and map_half_B)

    • Note: Both half-maps should be present in the input volume dataset; otherwise, if only the full map is present, signal from the full map will be introduced into each of the subtracted particles, which may result in particle half-sets that are no longer fully independent

  • Mask

    • Note: The mask should cover the region of the volume that is to be projected and subtracted out from the particle images (for more information on generating such a mask, see the Mask Selection and Generation in UCSF Chimera page)

Mask Troubleshooting

Note: Mask padding and size are among the most common sources of difficulty when using local refinement and particle subtraction. Please ensure that the mask is softly padded before running particle subtraction.

It is very important that the mask has soft padding, i.e., that the edge of the mask falls off gradually over space from 1 to 0. Masks can be padded using the Volume Tools job, by setting the Threshold and Soft padding width parameters. We typically recommend setting the Soft padding widthparameter to be a voxel value that ensures the resulting padding is several factors larger than the expected resolution of the complex. A good rule of thumb gives the minimum mask softness as 5 * res_angstroms / pixel_size_angstroms.For example, a complex reaching 3.0 Å in resolution with a pixel size of 1.0 Å should be padded to at least 5 * 3.0 / 1.0 = 15 voxels.

Common Parameters

  • Number of particles to subtract: The number of signal-subtracted particles to output. Runtime will be faster for fewer particles. If left as None, signal subtraction will be done to all particles

  • Low-pass Filter Input Structure (Å): Resolution (in Angstroms) corresponding to the low-pass filter applied to the volume


  • Particles (with signal subtraction done)

Example Images:

Useful Notes and Limitations


  • Masking: the mask is used the select the region of the volume that will be subtracted out of the images. As such, it might be a good idea in general to keep the mask somewhat tight and exclude any of the unresolved signal around the structure. Soft edges are recommended to reduce the Fourier-space ringing that may be caused by a sharp cut through the structure.

  • Windowing and Scaling: The particle window inner and outer radii parameters of a Particle Subtraction job must match the window radii used by the refinement job that produced the consensus refinement. The wrong windowing parameters may cause the subtracted volume to be at the wrong multiplicative scale, which will result in imperfect signal subtraction.

    • Note that the outputted particles will not be windowed, and this parameter is only used to calculate the required scaling factors.

  • Particles: the particles you feed into a Particle Subtraction job must already have alignment information - that means that they must have been through a refinement job prior to being imported. Particles from Local Refinement jobs can also be used, although a typical workflow would involve doing particle subtraction prior to a local refinement.

  • Gold Standard: the particles and volumes inputted to this job must be from a previous refinement (and not directly from an Ab-Initio reconstruction), as they must contain the two independent half-maps and the split of the particles. This job uses this split to independently subtract each half-map only from the particles used to generate it. This prevents the subtraction from introducing signal from one half-map into the other in future refinements.

  • Disk Space and Outputs: The particle subtraction job will create outputs in its job directory containing an full particle stack which will be approximately the same size as the input particle stack. The particle data files (.mrc) will be outputted in smaller batches corresponding to the memory available in the GPU, but a single .cs file will contain the metadata for the entire stack. Make sure you have enough space on disk before you run this job!

  • SSD Cache: Under Compute Settings, setting Cache particle images to SSD as true will store particle images on the SSD for faster processing. This requires adequate space on the SSD, the amount of which is determined at runtime.


  • For molecules with significant flexibility, Particle Subtraction is only an approximation of the true signal subtracted images, due to the fact that particle poses are assumed constant and are not iteratively refined as in Local Refinement.

  • Currently, Particle Subtraction supports volumes/masks that are sampled with the same box size as the incoming particles

Common Next Steps:

  • Local Refinement for further refinement of the sub-volume (whose signal remains in the subtracted particles)