Reference motion correction trajectories

Motion Correction
1

Hello,

I want to generate some histograms of the total particle displacement (full frame + per particle) in a dataset. Any recommendations on what fields to pull this data from? I see there is a particles.motion field, does this already integrate the full-frame + patch + particle motion trajectories?

Thanks,
-Craig

2

Hi @yoshiokc. Yes, the motion/path field contains paths to a numpy array, which itself has the rigid + RBMC trajectories. To get the RBMC-only path, you’ll have to load and subtract the RBMC path from this total path. Note also that in the CryoSPARC GUI (but not the data you download here), these trajectories are scaled by a factor of 40 so that they’re visible on the micrograph scale.

Here’s a cryosparc-tools script to help you do this:

from cryosparc.tools import CryoSPARC
import json
import numpy as np
from pathlib import Path
import matplotlib.pyplot as plt

with open(Path('~/instance-info.json').expanduser(), 'r') as f:
    instance_info = json.load(f)

cs = CryoSPARC(**instance_info)
assert cs.test_connection()

project_uid = "P312"
job_uid = "J142"

project = cs.find_project(project_uid)
rbmc = project.find_job(job_uid)
particles_rbmc = rbmc.load_output("particles_0")

pmc_juids = np.unique([x.split('/')[0] for x in particles_rbmc["location/micrograph_path"]])
pmc_micrographs = {puid: project.find_job(puid).load_output("micrographs") for puid in pmc_juids}

downloaded_motion_files = {}

def get_trajectories(particle_uid):

    p_rbmc = particles_rbmc.query({"uid": particle_uid})[0]
    total_motion_filename = p_rbmc["motion/path"]

    if total_motion_filename not in downloaded_motion_files:
        downloaded_motion_files[total_motion_filename] = np.load(project.download_file(total_motion_filename))

    total_trajectory = downloaded_motion_files[total_motion_filename][p_rbmc["motion/idx"]]

    mic_uid = p_rbmc["location/micrograph_uid"]
    pmc_juid = p_rbmc["location/micrograph_path"].split("/")[0]
    pmc_mic = pmc_micrographs[pmc_juid].query({"uid": mic_uid})[0]
    rigid_motion_filename = pmc_mic["rigid_motion/path"]

    if rigid_motion_filename not in downloaded_motion_files:
        downloaded_motion_files[rigid_motion_filename] = np.load(project.download_file(rigid_motion_filename))

    rigid_trajectory = np.squeeze(downloaded_motion_files[rigid_motion_filename])
    rbmc_trajectory = total_trajectory - rigid_trajectory

    return total_trajectory, rigid_trajectory, rbmc_trajectory

So, for example:

uid = 14502624203646355101
total, rigid, rbmc = get_trajectories(uid)

fig, axs = plt.subplots(1, 4, figsize = (16, 3), frameon = False, layout = "constrained")

# I have no idea why the forum is coloring these blue??
axs[0].plot(total[:,0], total[:,1])
axs[3].plot(total[:,0], total[:,1])

axs[1].plot(rigid[:,0], rigid[:,1])
axs[3].plot(rigid[:,0], rigid[:,1])

axs[2].plot(rbmc[:,0], rbmc[:,1])
axs[3].plot(rbmc[:,0], rbmc[:,1])

Makes this plot:

image
image1611Ă—311 35.6 KB

3 Likes
3

Thanks Rich!

I took a look at some of these values, just want to confirm they are the fitted “center” position of the particle in each frame and not a frame-to-frame shift vector. So I’d want to do something like:

def distance_traveled(trajectory):
    return np.sum(np.linalg(trajectory[1:]-trajectory[:-1], axis=1))

to get the total distance travelled?

Out of further curiosity, rigid here looks like it reflects the full frame motion, but not the patch motion for the particle neighborhood? does that mean RBMC ignores the patch motions when it begins fitting the particle trajectories? or does it just initialize the particle trajectory based on the patch?

4

Right – that would be the total distance traveled in pixels (assuming you meant np.linalg.norm() :wink:).

And right again, RBMC takes the rigid motion estimates from PMC but discards the motion of the patches themselves.

5

thanks!, yeah was a typo leaving out .norm,
but you then pre-emptively answered my next question!, Ă… vs. pixels. nice!