Yuanxin Zhu

Automatic particle detection through efficient Hough transforms

Manual selection of single particles in images acquired using cryo-electron microscopy (cryoEM) will become a significant bottleneck when a very large number of images are required to achieve three-dimensional reconstructions at near atomic resolution. Investigation of fast, accurate approaches for automatic particle detection has become one of the current challenges in the cryoEM community. At the same time, the investigation is hampered by the fact that few benchmark particles or image datasets exist in the community. The unavailability of such data makes it difficult to evaluate newly developed algorithms and to leverage expertise from other disciplines. The paper presents our recent contribution to this effort. It also describes our newly developed computational framework for particle detection, through the application of edge detection and a sequence of ordered Hough transforms. Experimental results using keyhole limpet hemocyanin (KLH) as a model particle are very promising. In addition, it introduces a newly established web site, designed to support the investigation of automatic particle detection by providing an annotated image dataset of KLH available to the general scientific community.

Rapid routine structure determination of macromolecular assemblies using electron microscopy: current progress and further challenges.

Although the methodology of molecular microscopy has enormous potential, it is time consuming and labor intensive. The techniques required to produce a three-dimensional (3D) electron density map of a macromolecular structure normally require manual operation of an electron microscope by a skilled operator and manual supervision of the sometimes complex software needed for analysis and calculation of 3D maps. Systems to automate the process of data acquisition from an electron microscope are being developing and these systems are being integrated with specimen handling operations and post acquisition data processing. Here, the current performance of our existing systems and the future challenges involved in substantially improving both the sustained throughput and the yield of automated data collection and analysis are reported.

Fast detection of generic biological particles in cryo-EM images through efficient Hough transforms

One of the current challenges in the analysis of a very large number of images acquired using cryo-electron microscopy is to investigate fast, accurate approaches for automatic detection of biological particles. Cross-correlation with a reference image based techniques have been developed in the past for this purpose; both computational requirements and accuracy, however, have limited their applicability to only a very small set of particles. The paper describes a new computational framework for fast, automatic particle detection, through the application of edge detection and a sequence of ordered Hough transforms. In particular, it presents how to adapt the generalized Hough transform to efficiently detect approximately rectangular shapes in a cluttered background. Preliminary results using hemocyanin as a model particle are promising.

Contaminant detection: improving template matching based particle selection for cryoelectron microscopy

Cryoelectron microscopy (cryoEM) has emerged as a powerful technique for structure determination of macromolecular assemblies (particles) in cell biology and is in great need of automation because it is both labor-intensive and time-consuming. One of the key steps in the automation is the selection of large numbers of individual particles in the cryoEM-acquired images. Due to the fact that images of this type usually have a very low signal-to-noise ratio, the current prevailing methods for particle selection are based on template matching, namely searching for peaks in the space formed by cross-correlating one or multiple reference templates with an entire image. Such methods generally require a second stage of human screening to remove false positives due to contaminants. This paper presents advancement in eliminating the human screening through automatic contaminant detection. Results of experimenting with both ribosome and GroEL particles demonstrate that on average the false positive rates can be reduced by about 6% using the proposed approach.

Automated cryo-electron microscopy

Cryo-electron microscopy is widely viewed as a uniquely powerful method for the study of membrane proteins and large macromolecular complexes - subjects that are viewed as extremely challenging or impossible to study using x-ray or NMR methods. Although the methodology of molecular microscopy has enormous potential, it is time consuming and labor intensive. Our group has done extensive work to automate image acquisition and processing for cryo-EM. In this paper we will provide an overview of our automated system, called Leginon, and present results where we used tobacco mosaic virus (TMV) as a proof of concept.