PyTME (Python Template Matching Engine): A fast, flexible, and multi-purpose template matching library for cryogenic electron microscopy data

Abstract

Cryogenic electron microscopy (cryo-EM) is a key method in structural and cell biology. Analysis of cryo-EM images requires interpretation of noisy, low-resolution densities, which relies on identifying the most probable orientation of macromolecules in a target using template matching. Many method-specific template-matching software solutions exist for single-particle cryo-EM, cryo-electron tomography (cryo-ET), or fitting atomic structures into averaged 3D maps of macromolecules. Here, we report the Python Template Matching Engine (pyTME), a software engine that consolidates method-specific template matching problems. The underlying library provides abstract data structures for storing and manipulating input and output data. PyTME runs up to ten times faster without loss in accuracy compared to existing software with multiple CPUs and GPUs, enabling template matching of even unbinned cryo-ET data in hours, which was previously nearly impossible due to technical constraints. Any hardware-specific optimization needed for dealing with large data is automatically performed to increase ease of use and minimize user intervention. The efficiency and simplicity of pyTME will enable high throughput mining of a variety of cryo-EM and ET datasets in the future.