Tutorial: Blob Picker Tuner

Optimize blob picks with the Blob Picker Tuner.

The Blob Picker Tuner works by comparing Blob Picker picks across different sizes, shapes, and thresholds to manual picks. It can take some of the guesswork out of using the Blob Picker job by quantitatively comparing sets of parameters. In this tutorial, we will see how to get good manual picks and set up the Blob Picker Tuner job.

Getting Manual Picks

Getting good manual picks is important for the blob tuner to perform well. The key point is to inform the grid search about the size of the particles by making picks close together:

It is also beneficial to choose the highest quality micrographs in the dataset for manual picking. Picking 15-40 particles is the recommended amount, but using more may improve the final result.

Running the Job

Connect the CTF-corrected micrographs to the micrographs input group. Connect the manually picked particles to the particles input group. The default parameters should be sufficient for most datasets, but the particle agreement distance must be set. This parameter controls the distance at which two picks are considered the same, and is important for this job to run correctly. The particle agreement distance should be roughly the size of the particle in Angstroms.

The parameters in the Template Parameters section are the same as in the Blob Picker job, and are not varied in the grid search.

When the blob picker finds more than 5 sets of parameters with the same best score, it will suggest labelling more particles.

The best set of parameters is printed out to the console for use later.

The outputs can be connected to an Inspect Picks job to verify the quality of the picks.