Examples of reconstructions
Outside of the training dataset - EMPIAR 12262
CryoLithe is a pre-trained model and yet it produces accurate reconstruction on new unseen dataset. Here, we apply CryoLithe to tilt-series from EMPIAR-12262, which contains non-infected cos-7 cells sampled at 5.525 Å/pixel. Crowded cellular environments are challenging to reconstruct accurately, a lot of overlapping features are projected onto the tilt series, and it becomes difficult to disentangle them.
Tips: Use the mouse wheel to zoom in and out at specific locations. The buttons below allows to zoom in and out from the central pixel. The slice slider allows you to chose the z-slice to visualize.
Quantitative analysis - EMPIAR 11830
We quantitatively validate the performance of CryoLithe on tilt series not seen during training, but obtained from the same dataset (EMPIAR 11830). These test tilt series were specifically selected to ensure strong performance of the baseline pipeline (Icecream). CryoLithe achieves this performance with a significant speed advantage, up to 75x faster, and without any parameter tuning.
Impact of CTF - EMPIAR-11830
To evaluate the impact of CTF correction on tomographic reconstruction, we use a tomogram of Chlamydomonas reinhardtii thylakoid membranes obtained from EMPIAR-11830 (test dataset). CryoLithe performs similarly when given the CTF corrected tilt series and the non-CTF corrected tilt series.
