Picking

EPicker is designed to pick general biological objects shown in cryoEM micrographs, currently, suports picking of particles, vesicles and fibers.

Picking command

The training script is epicker.sh in Epicker/Epicker/bin. Run the command with parameters

$ epicker.sh {Parameters}

Run the program without following any parameters to get the help infomation.

$ epicker.sh

Essential Parameters

Set the input and output path, select the model and picking mode using the following essential parameters.

• --data Micrographs to be picked, which can be a mrc file, a path to mrc files or a micrograph-list .thi file.

• --load_model Pre-trained model.

General models for picking particles, liposomes and filaments are available at Download.

• --mode (default=particle) Picking mode: particle, vesicle or fiber.

• --output Folder to output coordinate files.

Optional parameters

These parameters are useful and may influence the final results. If not spcified, the default setting will be used.

• --K (default=2000) Max particles to pick in each micrograph.

It is recommand to have a coarse estimation for the number of particles in one micrograph in average. Set the value to slightly larger (such as 20% more) than the estimated number.

• --vis_thresh (default=0) Threshold of matching score for particles to be picked. The value range is from 0.0 to 1.0.

The better matching to the learned features, the higher score. --vis_thresh is usually used with --K to accuately find the good particles in mcirographs. Empirically, 0.1 is good for most datasets, 0.05 or even smaller for dense particles in micrographs.

• --edge (default=25) Width of borders of micrographs to be excluded in picking.

Notice that the value refers to a width on the downsampled micrograph whose width(along x-axis) has been downscaled to 1024.

• --min_distance (default=0) Minimum distance between two particles.

In most cases, the default value works well. If many particles are picked multiple times, set a larger value to improve.

• --ang_thresh (default=0.3) Maximum curvature of tracing fiber, ranging from 0.0 to 1.57 in radian.

This option is ONLY used for fiber mode. The fiber may be curved. The angle threshold is used to terminate the tracing.

• --visual The picking results can be stored inPNG format for quick screening by adding --visual to the parameters.

• --output_type (default=thi) Format of the coordinate file. For visicle and fiber mode, only .thi format is supported. For particle mode, more formats including star, box, coord are supported.

• --gpus (default=0) GPU device ID to be used.

The picking is now very fast, and thus only single GPU was supported.