Francesca Cavallo
Development and integration of a neural network to detect image misalignment in cryo-electron microscopy.
Rel. Massimo Salvi, Carlos Oscar Sánchez Sorzano, David Herreros Calero. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Biomedica, 2026
|
Preview |
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (6MB) | Preview |
Abstract
In cryo-EM, particle misalignment —- arising from factors such as preferred orientation, low signal-to-noise ratio, or poor initial models —- affects the quality of 3D reconstruction and refinement. While several deep learning tools have improved map resolution, map anisotropy, pose estimation accuracy, or have incorporated quality control modules as part of the pipeline, no dedicated validation tool has been proposed to detect particle misalignment independently of its cause. The present study proposes cryoCheck, a supervised deep learning approach designed to be used as a validation tool to detect particle misalignment in the cryo-EM pipeline. CryoCheck employs an MLP architecture trained on Fourier Ring Correlation (FRC) curves computed between simulated projections and their corresponding experimental images.
Once the pose parameters of each cryo-EM image in the dataset are extracted, they are used to generate an aligned projection of a reference 3D density map, as well as a misaligned projection obtained by perturbing these parameters
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
