Inaccurate gain correction

Motion Correction
1

Hello,

We recently upgraded the camera on our Krios to Direct Electron Apollo, and it appears that cryoSPARC (v 4.4) has some difficulty with gain correction in the process of motion correction. The attached image contains the gain (estimated from the data to eliminate the possibility of the camera software introducing unwanted rotations/flips), output of motion correction in relion and output of motion correction in cryoSPARC for the same movie. Because the movies are in tiff format, the gain was “flipped in Y” when importing the movies into cryoSPARC. This approach worked well for our old Gatan K2, but for the Apollo gain correction in cryoSPARC produces artifacts. Just in case, I experimented with various combinations of rotations and flips, but the results were clearly worse. What could be the reason fro this?

Thank you.

gain-01
gain-011280Ă—720 164 KB

1 Like
2

Hi @tsybovsky

Are you using a defect file or are there any expected defects in the detector? Additionally, would you be able to provide us with a couple movies and the gain file you are using?

Best,
Kye

3

Hi and thank you for your response.

Since the initial post, I realized that the problem is likely camera defects. Relion’s motion correction is a CPU implementation of MotionCor2, and so it should detect hot pixels automatically and try to make corrections. My understanding is that cryoSPARC does not do it, or at least needs a defect file to do it, and this explains the different outcomes. Is this correct? Having a definitive answer would help as I am talking to the camera manufacturer. Below is a Google Drive link to the gain file and several movies. Thanks a lot.

4

Hi @tsybovsky

Thanks for sending over a link to those mics, we will take a look and get back to you. There is a possibility that the hot pixel detection is different between CS and RELION. We detect hot pixels by a value which is 5 standard deviations above or below the mean for the micrograph, and replace them with Gaussian noise that has the same mean and standard deviation as the micrograph.

Best,
Kye

5

Hi @tsybovsky,

We have been able to replicate the issue and are investigating. Thanks for sending us your data, much appreciated.

Best,
Kye