Abstract
Tubular structures are ubiquitous in biological systems and have been a focal point of cryo-electron microscopy (cryo-EM) structural analysis since the technique's inception. A critical step in processing tubular cryo-EM data is particle classification by diameter, as uniformity in diameter is a prerequisite for high-resolution three-dimensional reconstructions. Conventional methods rely on cross-correlation-based algorithms, which require prior knowledge to generate reference images, or iterative two-dimensional (2D) classification, that align and cluster particles into a predefined number of classes-a process that is both time-consuming and subjective. To address these limitations, we developed diameTR, a computational tool that rapidly determines tubular diameters in a prior knowledge-free and reference-free manner using GPU-accelerated processing on a per-particle basis. When applied to homogeneous datasets, diameTR yields narrow diameter distributions aligning closely with published values. For heterogeneous samples, it enables the separation of subsets with distinct diameters, validated by 2D averaging. Notably, diameTR identified an unreported smaller diameter subset of particles with new helical symmetry parameters in the previously published KpFtsZ-Monobody dataset. Collectively, diameTR represents a robust, efficient solution for diameter determination in tubular cryo-EM samples, eliminating the need for extensive human intervention while significantly accelerating processing.