Membrain-Pick

Overview

Membrain-Pick is part of the MemBrain suite of tools for processing membranes in cryo-electron tomography. MemBrain-pick's main purpose is to localize membrane-associated particles in the tomograms. To this end, MemBrain-picks takes as input already existing membrane segmentations and processes these to limit the search space for membrane-associated particles. The output of MemBrain-pick is a set of coordinates that can be used for further analysis.

Use case

MemBrain-pick is most powerfully used when training data is still limited or still needs to be generated. By leveraging membrane segmentations, MemBrain-pick can focus the search for particles to the membrane vicinity, reducing the amount of data needed for training.

The inherently used data representation corresponds to the one also used in surforama , allowing for seamless integration between both tools. MemBrain-pick can be used to generate training data for membrane-associated particles, which can then be manually annotated in surforama. The trained model can then be used to predict particle locations in new tomograms.

MemBrain-pick is therefore also limited by what is visible in the surforama interface. For instance, if particles are not clearly visible in surforama, MemBrain-pick will also struggle to localize them. It cannot recover structures that are not visible to a human annotator A general recommendation is therefore to first inspect the membrane segmentations in surforama before training a first model.

Workflow

The workflow of MemBrain-pick is as follows:

1. Input: Membrane segmentations in the form of a binary mask (.mrc). Ideally, these segmentations should depict single membrane instances.
2. Mesh Generation: The membrane segmentations are converted into a mesh representation. At this stage, also tomogram densities are projected onto the membrane mesh.
3. Ground Truth Generation: The membrane mesh can be loaded into surforama to manually annotate membrane-associated particles. These annotations can then be used to train a MemBrain-pick model.
4. Training: The generated meshes, along with the annotations, are used to train a model that can predict the location of membrane-associated particles.
5. Prediction: The trained model is used to predict the location of membrane-associated particles in the membrane segmentations.
6. (Optional) Subtomogram averaging: The predicted particle locations and initial orientations can be used for subtomogram averaging. Here, we explain how to import them into RELION for this purpose.

Example notebook

For a quick start, you can walk through our example notebook. You can easily run it on Google Colab by clicking on the badge below:

Key Functionalities

• Mesh Conversion: Transform membrane segmentations into a mesh representation that can easily be processed by MemBrain-pick and surforama.
• Model training: Train a model to predict the location of membrane-associated particles.
• Prediction: Use the trained model to predict the location of membrane-associated particles in membrane segmentations.
• Initial orientaton assignment: Given a set of positions, MemBrain-pick can assign initial orientations to the particles by aligning them with the membrane normal. -- integration with surforama: MemBrain-pick can be used in conjunction with surforama to manually annotate membrane-associated particles.

---

Jump to

• Installation
• Data Preparation
• Training
• Prediction
• Angle Assignment
• Visualization in Surforama
• Importing into RELION-5

MemBrain-pick is part of the MemBrain v2 [1] package and still under early development. If you have any questions or suggestions, please raise an issue on GitHub or contact the authors.

[1] Lamm, L., Zufferey, S., Righetto, R.D., Wietrzynski, W., Yamauchi, K.A., Burt, A., Liu, Y., Zhang, H., Martinez-Sanchez, A., Ziegler, S., Isensee, F., Schnabel, J.A., Engel, B.D., and Peng, T, 2024. MemBrain v2: an end-to-end tool for the analysis of membranes in cryo-electron tomography. bioRxiv, https://doi.org/10.1101/2024.01.05.574336