GPCR

Dataset

Learning objectives

In this case study, we'll look into the advantages of non-uniform refinement and how it can be used to improve the resolution of GPCRs and other classes of targets such as smaller protein complexes.

Title Value
Title
Description Leave empty
Directory
Create Workspace Enable

Parameter Value
Movies data path
Gain reference path
Flip gain ref & defect file in Y? Yes
Raw pixel size (A)
Accelerating Voltage (kV)
Spherical Aberration (mm)
Total exposure dose (e/A^2)

This should take less than 10 seconds to run.

Parameter Value
Number of GPUs to parallelize

Parameter Value
Number of GPUs to parallelize

Parameter Value
Minimum particle diameter (A)
Maximum particle diameter (A)
Use elliptical blob True

Inspect Picks outputs filtered pick locations so it's easier to compare and visualize different sets of picks at various thresholds. When you're happy with the set of particle locations, continue to the extraction phase to generated a set of particle images (a particle stack).

Select picks with an NCC >= 0.2 and Power Score > 635 and < 1045.

You should have around 134K pick locations.

Parameter Value
Number of CPU cores
Extraction box size (pix)
Fourier-crop to box size (pix)

Parameter Value
Number of GPUs to parallelize

Choose a set of 2D classes that look sharp and display secondary structure features.

You should have around 67K particles.

Use default parameters

Let's queue a Homogeneous Refinement to compare it to the Non-Uniform Refinement we'll run in parallel. Use all particles and the volume from ab-initio as inputs.

Parameter Value
Minimize over per-particle scale Yes

Launch this job in parallel with the previous Homogeneous Refinement. Use the same particles and volume from ab-initio as inputs.

Parameter Value
Minimize over per-particle scale Yes

Final Result

You should have a 3D reconstruction of the GPCR at ~3.6A resolution.