Markerfree: GPU-accelerated marker-free alignment for improved cryo-ET reconstruction

Tilt series alignment is pivotal for cryo-electron tomography (cryo-ET). However, the increasing diversity of cryo-ET datasets, combined with challenges such as low signal-to-noise ratios and complex cellular structures, has significantly raised the demands for tilt series alignment. Although marker-based alignment provides greater precision and robustness, it is not always feasible, which increases the need for marker-free alignment. However, current methods and software often fall short in delivering the accuracy and robustness required for certain datasets. Here, we introduce Markerfree, a GPU-accelerated software package for fully automated and robust tilt series alignment. The pipeline integrates iterative cross-correlation and common-lines-based coarse alignment, followed by global refinement using enhanced projection matching, which limits the number of images participating in reconstruction while improving their quality to achieve superior precision. Analyses of both simulated and real-world datasets validate Markerfree's accuracy and robustness, positioning it as a powerful tool for modern cryo-ET workflows.