ClassClean

Use the ClassClean logic to separate good particle images from bad “junk” particles, a general prerequisite for obtaining a good 3d structure.

The ClassClean logic computes a quality score for particles in context of a group of particles (“class”). Therefore, this logic requires particles to be aligned and classified beforehand - either by assignment to a reference from Alignment or a class from Classification. ClassClean can be operated in two modes: In “clean” mode particles are evaluated within each class and dependent on chosen parameters a certain fraction of only the best particles is kept. In “refine” mode existing classes from a previous “clean” mode run may be supplemented with a new set of particles. These new particles are then i) assigned to the best matching class, ii) evaluated within this class and iii) either kept when they improve the quality measure of this class or discarded. Using a circular mask helps in particle evaluation by removing background noise. Note: The ClassClean logic performs no alignment.

Use this mode to remove bad particle images from a data set.

Parameters Description
Class truncation - enabled Use this option to define how many of the best particles should be kept per class at maximum
→ Class truncation threshold Maximum number of best particles to be kept for each class
Class truncation - disabled Use this option to define the fraction of best particles to be kept solely by the “Sigma value”
Header key Define whether classes should be defined by preceding alignment (“referenceData”) or classification (“ClusterMember”)
Mask usage Define whether no mask (“none”), a circular mask or a soft circular mask should be used.
→ Relative radius Define radius of the circular mask
Minimum class size Minimum number of particles to be kept within each class.
Sigma value Define in terms of standard deviations what fraction of the best particles should be kept.
Input Description
Input Images Particle images to be cleaned. Note: This particle images must have been aligned and/or classified
Output Description
Class members Good particle images to be kept
Class sums Class averages from cleaned classes containing only the good particles
Discarded Images Bad particles images discarded from the data
New/Changed Header Values Description
ClassCleanClassQuality Overall quality measure of the class the particle image belongs to
ClassCleanImageQuality Quality measure of the individual particle image within the context of its class
classSize Number of class members after cleaning

Use this mode to add only good particle images to an already existing “cleaned” data set.

Parameters Description
Header key Define whether classes should be defined by preceding alignment (“referenceData”) or classification (“ClusterMember”)
Mask usage Define whether no mask (“none”), a circular mask or a soft circular mask should be used.
→ Relative radius Define radius (in fractions) of the circular mask
Input Description
Input data (Preclassified)
Input data (Unclassified)
Output Description
Class members Good particle images to be kept
Class sums Class averages from cleaned classes containing only the good particles
Discarded Images Bad particles images discarded from the data
New/Changed Header Values Description
ClassCleanClassQuality Overall quality measure of the class the particle image belongs to
ClassCleanImageQuality Quality measure of the individual particle image within the context of its class
classSize Number of class members after cleaning
referenceData/clusterMember(depending on user selection) Class where the good “unclassified” images have been placed by ClassClean