This paper introduces a new Bayesian approach for subtomogram averaging that operates directly on 2D tilt series images within RELION 4.0. This method eliminates the need for missing wedge correction and offers improved accuracy for subtomogram alignment and classification, demonstrated by achieving resolutions sufficient for de novo atomic modeling in two test cases: the HIV immature capsid and the C. crescentus S-layer.

None identified

This paper presents a significant methodological advance in cryo-ET image processing. The shift to a 2D likelihood-based approach addresses limitations of previous 3D methods. The implementation in RELION 4.0 makes this powerful technique accessible to a wider audience. While the evaluation is limited to two case studies, the results are promising. The lack of a detailed discussion of limitations and a broader comparison with other software packages prevent a rating of 5.