I used template picking on 100 micrographs and found ~93k particles. Then I extracted these particles, using the template job’s particles and micrographs as the inputs. The extraction ends up with only ~70k particles after “intersection with the passthrough dataset.”
Is there a closeness threshold or duplicate picks remover operating in the background? Or another reason? Thanks.
Particles that are too close to the edges of the micrographs (close enough to cause the extracted box to extend beyond the the micrograph) are discarded. Though losing a quarter of the particles seems a little too much - are you using a very large box size compared to the micrograph size? What happens if you try to re-extract with a much smaller box size?
I hope that helps, please let us know what happens.
These are K2 images with a pixel size of 1.2156, similar to EMPIAR-10059 (but apo). The exact numbers were 83,890 from template picking, and from extraction 72,159 (288 px box) or 76,211 (192 px box), so it’s more like 15% that are lost. TRPV1 can be refined well with a much smaller box, but the large box is theoretically necessary to recover all of the signal from the high defocus images, and the performance hit is essentially inconsequential.