Quick Start: Tilt Series

Tilt series pre-processing in Warp will allow you to reach higher resolution for your sub-tomogram averages. However, unlike the pipeline for 2D data, the tilt series pipeline still has a few gaps. You will have to switch to IMOD to perform the initial tilt series alignment (using patch tracking or gold fiducials). There is also no BoxNet-enabled on-the-fly particle picking yet, and you will have to use a rather slow template matching routine in offline mode. Metadata import is supported only for SerialEM currently. We intend to address these shortcomings in the near future.

We will use data from EMPIAR-10164 in the screenshots below – in particular, the 5 tomograms used in Turonova et al. 2017.


Tilt movie pre-processing

The first step towards tilt series pre-processing in Warp is pre-processing the individual tilt movies in 2D mode. If you’re not familiar with that procedure or Warp’s basic interface features yet, please follow the 2D quick start guide. Here are some important differences to traditional 2D movie processing for this step:

The motion model shouldn’t have spatial resolution, as there is usually not enough signal in a single tilt movie average for such fine alignment. Thus, the grid resolution will be 1x1xN. A good value for N is the number of frames in the shortest movie in case you’re using a collection scheme that increases the dose (and thus the movie duration) at higher tilts. If you’re fractionating the dose in very fine steps and collecting with a small pixel size, N can be lower than the frame count to avoid overfitting.

The CTF can usually be estimated well with a 2x2x1 defocus grid. You can use these estimates to monitor data collection. However, the values will be re-estimated later during tilt series pre-processing using a more robust model.

Particle positions obtained with BoxNet won’t be considered later. However, the masks will be. In tilt series mode, those are assumed to cover gold beads and other high-contrast artifacts, and will be used to fill image areas with zeros. You can quickly retrain BoxNet to mask out gold beads and not pick any particles. We would like to offer a pre-trained model for this purpose, but currently lack the training data. If you think such a model would be useful, please consider contributing training data!


Export tilt series stacks for IMOD

IMOD’s etomo requires each tilt series to be assembled in a single MRC stack. You can use Warp to combine the previously generated tilt movie averages into such stacks.

Switch Warp to tilt series mode by selecting TomoSTAR as the Input format. We haven’t created any TomoSTAR files yet, but this mode will allow us to generate the data necessary for tilt series alignment in etomo, which will eventually allow us to create TomoSTAR files.

Once in tilt series mode, the Overview tab will contain the Import tilt series from IMOD button. Click it to bring up the task dialog. Specify the folders containing the mdoc files generated by SerialEM, and the original tilt movies. If there are multiple mdoc files per tilt series (e. g. “plus” and “minus” after bi-directional collection), you can specify several suffixes. Once these settings are set, all tilt series should appear with the correct number of tilts in the list on the right. Individual series can be deselected using the checkboxes next to their names. Click the Create stacks for IMOD button to generate the stacks. This will create an imod folder in the current project directory, and place each stack in a sub-folder named after the respective tilt series. You can now execute etomo in each of these folders to perform alignments. Please note that it isn’t necessary to generate an aligned stack or reconstruct tomograms in etomo, other than for diagnostic purposes. You can use manual deselection in Warp’s 2D pre-processing mode to remove individual tilt images from the stack. Please remember to reflect these changes when setting up the project in etomo.

You can also generate TomoSTAR files without any alignments at this point by clicking the Import button. Please refer to the next section for information on how to specify the electron dose.


Create TomoSTAR files from IMOD’s results

If you closed the Import tilt series from IMOD dialog from the previous section, open it again. Specify the mdoc and original tilt movie folders, and the suffixes (if applicable) as before. For the Root folder with IMOD processing results, select the imod subfolder created previously. If you didn’t use Warp to set up the folder structure as suggested in the previous section, make sure the alignment results are located in [project folder]/imod/[tilt series name]/. Warp will try to find the xf and taSolution.log files there. If everything has been parsed correctly, the Aligned checkboxes should become checked.

Specify the same Pixel size you used in IMOD. Warp will use it to convert the shifts to Ångström. For the electron dose, there are two options: SerialEM, when calibrated accordingly, can store the dose values in the original mdoc files. If the calibration is off or hasn’t been done at all, you can specify a non-zero value in the Default dose per tilt field. There is currently no option to conveniently specify varying dose values. In case of dose-symmetric collection, it is usually OK to just use the per-tilt dose of the very first tilt images for all tilts, since the higher tilts won’t contribute much information to high-resolution 3D averages anyway. If you absolutely need to specify varying dose values, you will have to edit the wrpDose column in the TomoSTAR files later.

That’s it! Click Import to create the files and start processing them in Warp.