
Residue 1 of 3
Phe106
The published density left the ring's position open to interpretation. Ours shows it clearly. So we changed the side chain's rotamer followed by a small local refinement in Coot.
We re-picked the public dataset behind the published PfCRT structure, a 49 kDa parasite membrane protein bound to a ~50 kDa Fab, with our Topaz best practices: 5× more particles in the final 2.88 Å map, 9.4× more rare view particles, and clearer side chains. No new microscope time.
2.88 Å
Map resolution
3.3 Å published
5×
More particles
84,170 vs. 16,905
9.4×
More rare view particles
18,168 vs. 1,931
3
Side chains re‑fitted
Phe106, Asn167, Lys284
Mutations in PfCRT help malaria parasites resist drugs by moving chloroquine and piperaquine away from where these drugs normally act. Kim et al. (2019) provided the first structural view of this mechanism. Because PfCRT is small (49 kDa), accurate particle picking is critical for achieving high-resolution cryoEM structures. The published map used 16,905 particles from over one million picks, and some side chains remained ambiguous.
Our Approach
We re-picked particles beginning with the raw EMPIAR-10330 movies using the Topaz best practices taught in our Topaz Masterclass, then refined the structure in CryoSPARC.
Residue-Level Evidence
Choose a residue to compare the published density with ours.

Residue 1 of 3
The published density left the ring's position open to interpretation. Ours shows it clearly. So we changed the side chain's rotamer followed by a small local refinement in Coot.
Published vs. Reprocessed
Particle picking
Particles in the final map
Rare view particles
Resolution, PfCRT with the Fab variable domains (FSC = 0.143)
Refinement
| Measure | Published (Kim et al., 2019) | Our Reprocessing |
|---|---|---|
| Particle picking | DoG picker and Gautomatch | Topaz with our Masterclass best practices |
| Particles in the final map | 16,905 | 84,170 (5× more) |
| Rare view particles | 1,931 | 18,168 (9.4× more) |
| Resolution, PfCRT with the Fab variable domains (FSC = 0.143) | 3.3 Å (PfCRT at 3.2 Å) | 2.88 Å |
| Refinement | CryoSPARC non-uniform refinement, with CTF refinement in cisTEM. Nanodisc and Fab constant domain were then subtracted in RELION, followed by a final refinement. | CryoSPARC non-uniform, then local refinement on PfCRT and the Fab variable domains |
3.3 Å (the deposited map, EMD-20806) and 2.88 Å are measured over the same region, PfCRT with the Fab variable domains, at the same threshold.
CryoSPARC plots for the final local refinement, with a mask around PfCRT and the Fab variable domains (FSC = 0.143).
Methods & Data
Built on public data: every published number here is one click away. Open a panel for the full methods and analyses.
Raw movies from EMPIAR-10330 were motion-corrected with MotionCor3 in RELION, and a Topaz model was trained. The movies were then motion-corrected separately in CryoSPARC, where their CTFs were estimated and all further processing was done. Applying the Topaz best practices taught in our Topaz Masterclass gave 135,692 particles. Non-uniform refinement followed by local refinement with a soft mask around PfCRT and the Fab variable domains gave the final map: 2.88 Å resolution from 84,170 particles (gold-standard FSC0.143). Side-chain rotamers were re-fitted in Coot for Phe106, Asn167, and Lys284, which were ambiguous in the original map.
We split all viewing directions into 24 equal-area views. The six rarest views account for 1,931 of the published particles and 18,168 of ours: 9.4× as many, from 5× as many particles. Every view grew by at least 2.8×, and these six grew by the most.
Deposited Particles
16,905 particles (published angles)
Our Topaz Particles
84,170 particles
Both plots share the same color scale.
We also ran their deposited particles through our own processing and did not get better resolution than the published result. So in this case, the larger number of particles, not newer software, improved the resolution.
Other teams have used the same EMPIAR dataset to test new algorithms, reporting 3.6 Å (CryoSPARC non-uniform refinement; Punjani et al., Nature Methods 2020) and 3.4 Å (RELION Blush regularization; Kimanius et al., Nature Methods 2024). Both reported resolutions are for the whole particle, including the full Fab and the nanodisc, so they cannot be compared directly with the original 3.3 Å result.
Picking further with our Topaz best practices gave 114,378 particles, which refined to 2.95 Å, slightly short of the 2.88 Å from 84,170 particles, so we report on the 84,170-particle map. We suspect the extra particles are more mobile, or capture a slightly different conformational state of the protein. 3D Classification and 3DFlex on all three particle sets (the deposited 16,905, and our 84,170 and 114,378) gave inconclusive results that could neither confirm nor rule out this hypothesis.
PfCRT is a small membrane protein, about the size of many receptors and transporters in drug discovery. If your map is stuck at a resolution that's too low, or the side chains you care about are still ambiguous, talk to us before you book more microscope time. The answer may already be in your data. We scope and quote every project individually, and you work directly with the founders.
Processed in part using CryoSPARC. CryoSPARC™ and CryoSPARC Live™ are trademarks of Structura Biotechnology Inc. and are used under license.