ProtGautomatch
Gautomatch is a GPU accelerated program for accurate, fast, flexible and fully automatic particle picking from cryo-EM micrographs with or without templates, developed by Kai Zhang at MRC Laboratory of Molecular Biology. A predecessor of Gautomatch was developed in Dr. Fei Sun's lab 6 years ago. The basic concept was based on local cross-correlation (as described in FindEM paper) at that time. However, the concept and usage of newer version has been largely updated, which now requires only a single and simple command for accurate and 'smart' whole-dataset processing.
Brief manual describing program options and use cases together with some test data can be found on Kai's website.
To launch auto-picking you will need to first provide the Input micrographs, that can be any SetOfMicrographs produced earlier (Fig. 1). Then you have to select Input references, that can be any SetOfAverages: either 2D class averages, calculated before, or reprojections of the known 3D structure (in this case, you have to import your reprojections as Averages). Moreover, you can run the program without references - they will be automatically generated (this should work for spherical particles or ribosomes). If the input micrographs and references do not have the same contrast (e.g., white particles on a dark background), select the option References have inverted contrast. A good idea is to run picking on a subset of micrographs with different defocus values - for this you can use the corresponding option (you will need to provide a set of CTFs though).
Among other options the most important one is the Threshold: cross-correlation peaks only above this value will be selected as particles. Usually, you should first run the protocol on a few micrographs with different defocus level to optimize threshold and other parameters before starting auto-picking on a whole dataset.
The protocol provides interactive wizards for choosing the particle size and for optimizing threshold value. The last one allows you to adjust threshold and immediately display the results of the picking.
Fig. 1. Input tab of Gautomatch protocol GUI.
By default, the program will guess all advanced parameters by itself, however if you want to see all available options, choose the Advanced tab and select No for "Guess advanced parameters?" (Fig. 2).
Fig. 2. Switching off "guessing" parameters mode.
There are many more parameters to play with, for more detailed information you can click Help button. For example, if you are picking negatively stained particles, you might want to increase default value of local sigma cut-off (find this option yourself! :). If you are interested in picking a specific domain / orientation, have a look at exclusive picking options.
When the protocol is finished you may click on the Analyze Results button, it will ask you what particles you want to display: autopicked or rejected ones. The last option can be useful during parameter optimization. Another advice for expert users if you run the program the first time: on Debug tab choose to save CC and sigma files. They can be later opened e.g. in IMOD and with help of pixel viewer will give you an idea about useful threshold and sigma value settings.
Fig. 3. Displaying picking results.
- Zhang K. unpublished