Job: Blob Picker Tuner
Optimize blob picker parameters.
The Blob Picker Tuner finds optimal blob sizes, shapes, and thresholds for particle picking by comparing picks to a small number of manually selected examples. It does so by performing a grid search of the parameters available to the Blob Picker and Inspect Picks job and computing the agreement with the user picks. It then takes the optimal parameters and performs a full search on all of the input micrographs. Read a tutorial on the Blob Picker Tuner here.
Limitations: The Blob Picker Tuner can not discriminate particles better than the regular Blob Picker, but makes the process of identifying useful parameters much easier. For the Blob Picker Tuner, the number of labelled particles is important. Labelling too many particles may introduce bias into the Blob Picker based on the human bias of picking. On the other hand, labelling too few particles may result in poor picks. Somewhere from 15-40 particles picks is recommended, but more may be needed for difficult datasets. Neighbouring particles should be picked for manual input to ensure the tuner has information about particle spacing. Read more in the tutorial, here.
- Aligned and CTF-estimated micrographs
- Particle picks on one or more of the above micrographs
- Minimum template diameter (A): The minimum template size that the grid search will consider.
- Maximum template diameter (A): The maximum template size that the grid search will consider.
- Steps: The number of steps the search will take between the minimum and maximum diameters. More steps means a finer resolution in parameter space, with the cost of increased (quadratic) computation time.
- Particle agreement distance (A): Determines the distance at which a user-provided pick is considered the same as a blob-pick. The tuner works best when this parameter is roughly the size of the protein.
- 2D Classification and Select 2D to generate templates for Template picking