Lawrence Berkeley National Laboratory

Advances in Electron Microscopy and single-particle reconstruction have led to results at increasingly high resolutions. This has opened up the possibility of complete automation of single particle reconstruction. Main bottleneck in automation of single particle reconstruction is manual selection of particles in the micrograph. This paper describes a simple but efficient approach for segmentation of particle projections in the micrographs obtained using cryo-electron microscope. Changing the shape of objects to facilitate segmentation from the cluster and reconstructing its actual shape after isolation is successfully attempted. Both low-level and high-level processing techniques are used and the whole process is made automatic. Over 90 percent success in automatic particle picking is achieved. Several areas for improvement and future research directions are discussed.