Dose limits

Submitted by lukater on Tue, 2016-05-24 13:03

Forums

Unblur & Summovie

Hi everybody,

I have been trying to use the dose weighting of Summovie to improve my final reconstruction.
Initially I used sums without dosage correction, thereby I reach a final resolution of 3.3A. I then wanted to try the dosage correction.

For this I created new sums of my previously aligned movies. I used the par file of my 3.3A Volume and the particles from the new dose filtered sums, I end up with 4.24A, which improves to 3.95A after another 4 rounds of refinement. My movies consist of 10 frames with a dose of 3.1 e/A²/frame and a pre exposure of 4.6 e/A².

Do you have some input on the unexpected resolution difference?
Is there a dose range in which Summovie works optimally?

Best
Lukas

timgrant Tue, 2016-05-24 13:05

Hi Lukas,

How is your dose calibrated? Is it possible it is incorrect? What camera is this?

Cheers,

Tim

lukater Tue, 2016-05-24 15:00

The camera is a Falcon II. I had previously contacted the EM facility to ask about the accuracy of the dose as this was my first idea for a source of error. They informed me that the calibration is done via a faraday cage (but I have no further details on that process) and that they are fairly certain concerning the accuracy of the dose.

Cheers,
Lukas

timgrant Tue, 2016-05-24 15:22

Hi Lukas,

I've never seen that much of a difference between the filter and not - so my initial guess would be that your dose estimate is off, although I can't say there isn't another problem for sure. We typically use a counting detector, and then you can accurately estimate your dose from the images themselves. You could try to scan different doses to see if you can get a better result.

Other people have had this problem with Falcons, probably because the dose was incorrect - e.g. http://grigoriefflab.janelia.org/node/5035

One final thing - Is your sample all protein, or do you have DNA/RNA aswell, in which case the dose filter will not be optimal (although the difference you report is probably too large to simply be this).

Cheers,

Tim

lukater Tue, 2016-05-24 16:18

Hi Tim,

the sample is a 70S ribosome. I was not aware that this would be problematic.

Maybe I am also a bit confused about the dose filtering. What is the ideal outcome of a successful dose filtering? The way I understood the maps should appear sharper due to improved SNR at high frequencies. In this case, would the resolution not also improve?

Compared to the post you linked is seems my dose would then potentially be very far off my assumed values. Is there a way to make a rough estimate so I do not have to test a large number of settings?

Thank you very much for your help!

Best,
Lukas

timgrant Tue, 2016-05-24 17:12

Hi Lukas,

The exposure filtering is supposed to optimise the SNR at all spatial frequencies. This is something you can't normally do with a single sum - you can optimize the SNR at high frequencies by taking a sum of just your initial frames, but you will be using a lot of low frequencies that might be useful for the alignment. If you sum all your frames you get the low frequencies, but lose high frequencies due to radiation damage. I would say that given the 70S ribosome is such a high SNR object, you probably won't gain much using the dose filter over just using an sum of your (initial) frames, although it shouldn't be worse, at least for the protein.

Your 70S ribosome is composed of two materials, with different radiation sensitivity. It is basically impossible in a single reconstruction to optimise both the protein and RNA, the exposure filter was calculated on protein, so it won't be optimal for the RNA. The RNA tends to be the driver for the FSC so we've had ribosomes before where the FSC is worse (slightly) with the exposure filter, and the RNA looks worse, but the protein looks better.

I'm afraid I don't really have a good idea of how to work out your actual dose off the top of my head, other than trying half, 2/3rds etc and optimizing in from there.

Tim