Hi everyone,
I have a basic question. I’ve already collected a data set and processed it. I’ve found out that my sample has a preferred orientation.
Is there a way to process the same data set in a tilted mode or should I collect a new data set.
Thanks!
Collect a new dataset. There is no “tilted mode” of processing - it refers to literally tilting the stage and collecting data to get a broader range of orientations. I would also consider trying high-CMC surfactant additives to mitigate preferred orientation before trying tilting, as tilting will somewhat degrade resolution.
Having said that, depending how severe the preferred orientation is, you may be able to apply some processing tricks to squeeze more out of the data you already have!
Cheers
Oli
Thank you, Oli!
Great information.
What would be the tricks to squeeze more out of the data?
Depends on the data - you will need to share more about your specific case for folks to provide helpful suggestions
Thank you!
So I’ve been trying to solve the structure of a trimeric glycoprotein binding to three Fabs. So far the most dominant particle that I’ve been seeing after 2D classification is apparently a side view in which there are only two Fabs visible. There are similar particles with same orientation.
I was wondering if anyone has had the same problem.
Welcome to the forum.
Lots of people have the same problem with a wide range of samples. It’s pretty common. 
As @olibclarke has already mentioned, tilting the stage for an acquisition can solve it (but brings problems of its own), or trying surfactants to discourage interactions between the protein and the solvent/air interface, or changing the material of the grid, or using carbon support film, or changing glow discharge conditions, or changing blotting conditions, or… well, you get the idea!
Lots of things to try.
If you still have grids and the concentration isn’t too high (particles have empty ice around them) then a 20-30 degree tilt can be the quickest way to solve the problem, but if tilting gives particle overlap it’s a no-go. Also, depending on your 'scope/software, tilting may slow acquisition dramatically.
That said, if low concentration, I’d try tilting first, followed by carbon film, then high-CMC surfactant. If high concentration, I’d go straight to high-CMC surfactant.
Thank you very much!
So there are lots to try :))
I will design the experiments based on your recommendations.
agree that you likely have a common problem, many ways to overcome it with sample prep/data collection, but also agree that you might not have an inaccessible dataset. Depending on your willingness to share vs privacy, 1 micrograph, 1 picked micrograph, 2D of all particles would get you expert opinions about fastest path to structure.
Thank you!
Below is the 2D classes:
Box size : 512 Fourier crop: 64
Full iterations : 2 Online iterations: 40 Batchsize: 200 Force max 
Apparently I can’t post more photos since I’m a new user.
Only one of these classes looks convincingly like protein, and at first glance looks more like Fab than glycoprotein if I had to guess - perhaps sample denaturation/dissociation at the air water interface? In that case tilting will not help, and you should explore other alterations to sample prep
Box much too large, particles not correctly picked (try template picker with the one good class). Absolutely not possible to know if you have preferential orientation in ice.
This is another try for 2D classification:
Box size : 360 Fourier crop: 180 (The particle I’m looking for should be around 170 A)
Full iterations: 2 Online iterations:40 Batchsize: 200 Force max off
What are your thoughts?
Same thoughts as for the previous one - perhaps if you can post an example micrograph we might be able to give thoughts on picking & sample quality?
And perhaps the same micrograph with particle positions (e.g. form Inspect Picks)?
Looks quite overpicked, and a lot of picks on carbon. In your original image there are some particles visible that could be your trimers, though. I would recommend trying the micgrograph denoising tool to improve their contrast, then do some manual picking and 2D classification to see what you have. If you can get a decent set of ~1000 good manual picks, then train a Topaz or CrYOLO model to repick the entire set.
Thank you Oli!
I’m trying to see if I can denoise micrographs in cryoSPARC. Could you help me with that?
See instructions here:
Would recommend training your own model, which requires having run Patch Motion with 4.5 or later, but gives great results, e.g., here is an example of raw vs denoised for EMPIAR-11918 (a 40kDa protein-RNA complex):
Raw:
Denoised:
Great! Thank you!
Just another basic question. What is the best values for Power Score and NCC when I inspect pics? I’m trying to see the if I’m choosing the correct particles.
