Template matching and machine learning for cryo-electron tomography

Cryo-electron tomography is the best-suited imaging technique for visual proteomics. Recent advances have increased the number, quality, and resolution of tomograms. However, object detection is the bottleneck task of the analysis workflow because, so far, only a few molecules can be detected by computer methods for pattern recognition. This article introduces the major challenges in detecting molecular complexes for cryo-electron tomography. This paper also identifies the limitations of the current methods. Finally, it describes the approaches proposed to overcome these limitations.