Regarding pleomorphic enveloped viruses

[Ankushcreatedroid]

Can anybody forum suggest a good work flow to do uniform spike picking over a segemented pleomorphic virus geometry. Also if some body can give idea on how to segment pleomorphic virus to get there geometries out for picking uniform spikes over the membrane surface ?

Thank you
Ankush Roy

[abartesaghi]

For segmenting virions (spherical or pleomorphic), look at the Tomogram segmentation section in the user guide.

Once you have successfully segmented your virions, follow these steps to pick particles:

1. create a Particle picking (open surfaces, filaments) block (downstream of the segmentation block)
2. select the size of your particles (this is used for display purposes only)
3. choose between template search or uniform (and set the required parameters for the selected method)

This will allow you to run template search or pick uniform positions constrained to the segmented surfaces.

[abartesaghi]

We would need more information to troubleshoot this issue. We recently added an option that lets you switch between the original and denoised versions of each tomogram, what version of nextPYP are you using?

[Ankushcreatedroid]

i am currently updated nextPYP 0.7.3. Thank you will be waiting for a some sort of update before even running that big chunk of data for single denoise train.

[Ankushcreatedroid]

so u can see that i have succesfully trained model for 5 TS from batch of 270 TS but when i apply that trained model to whole batch then each position_*.mrc tomogram complains this message that in .hd5 file it can't find the file corresponding to the position.

[abartesaghi]

As I mentioned, we would need some additional details to properly troubleshoot this. Also, there’s no need to run cryoCARE denoising on the entire dataset right away. You could first select a few tilt-series as a test subset. Once you confirm that the results are as expected, you can proceed with processing the full dataset. This can be done easily in nextPYP using tilt-series filters. By the way, what settings are you using for training and what does the plot of the loss function look like for your training block?

[Ankushcreatedroid]

As you mentioned and also in tutorial i filtered some 5 TS from whole batch for denoising in this session. I got my expected result for surface protein to be more clear. Though my loss function oscillates around 6 as you see in screenshot. For the parameter i used particle box size of 72 default because i have arrays of particles on membrane surface plus training epochs of 100 and steps 200 as suggested in original cryoCARE website tutorial.

[Ankushcreatedroid]

To the previous suggestion you have given. I followed it and got good enough membrane segmentations of virus but now when i am doing uniform picking over segmented membrane why uniform picking is happening both inside and outside membrane and though i want only normals facing outwards. Any specific reason.

[Ankushcreatedroid]

This are my uniform picking parameters.

[abartesaghi]

What you are seeing is, unfortunately, the expected behavior. Because the membranes generated by MemBrain-seg are several voxels thick, the "outside" region of the membrane includes both an internal and an external surface. Each of these surfaces produces uniformly spaced positions.

That said, we do have newer tools that allow you to select only the outer surface. These tools have not yet been released, but I would be happy to share instructions if you email me directly.