Beyond Exhaustive Sampling: Efficient Rotational Matching via Ball Harmonics

Abstract: Cryo-ET allows to generate tomograms of biological samples in situ, capturing complex structures in their native context. Despite low signal-to-noise ratio in reconstructed volumes, the large number of copies of the same macromolecules makes it possible to retrieve high-resolution maps by averaging many aligned subtomograms. To keep up with technical advances in the imaging process and the resulting huge amounts of data available, there is a need for scalable, fast and robust procedures to align subtomograms. We propose a subtomogram alignment framework based on the ball harmonics expansion that combines frequency- and gradient-based optimization strategies to avoid exhaustive rotation sampling, enabling a speed-up of an order of magnitude compared to current approaches.