Negative results from two synthetic-injection studies of nanogold detection in chromatin electron tomography
We report two prospectively planned synthetic-injection studies of 1.4 nm nanogold detection in real tilt images from EMPIAR-12907. Both were negative. A fixed requirement combined recall of at least 0.80 within 1 nm with limits on spurious detections, accepted decoys and empty-control detections. Calibration covered uncertainty in diameter, defocus and per-view displacement. The independent units were three lamellae, probably from one preparation. In Study 1, none of 280 settings met the requirement in any of 39 held-out paired-seed cases. The calibration procedure abstained throughout; its formal failure-detection passes did not demonstrate usefulness. Post hoc analysis identified dense-object fragments and a separate recall ceiling. Study 2 tested an object-specific rejector and a contrast-transfer-aware template detector. It stopped at its development gate: no arm had a setting satisfying every uncertainty cell on any lamella. Rejection reduced fragments but lost nearby targets; the template implementation did not lift the recall ceiling. These results concern synthetic recovery under the tested model and preprocessing. They neither measure real-label accuracy nor evaluate the authors' chromatin pipeline. Real localisation remains on hold because alignment misses the stated detector requirement and independent annotation is absent.
Authors
- S. M. Abdulla Al Mamun
Institutions
- University of Manchester (GB)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-19
- DOI
- https://doi.org/10.5281/zenodo.24374081
- Primary Topic
- Advanced Electron Microscopy Techniques and Applications
- Type
- preprint