Automated screening of 2D crystallization trials using transmission electron microscopy: a high-throughput tool-chain for sample preparation and microscopic analysis.

We have built and extensively tested a tool-chain to prepare and screen two-dimensional crystals of membrane proteins by transmission electron microscopy (TEM) at room temperature. This automated process is an extension of a new procedure described recently that allows membrane protein 2D crystallization in parallel (Iacovache et al., 2010). The system includes a gantry robot that transfers and prepares the crystalline solutions on grids suitable for TEM analysis and an entirely automated microscope that can analyze 96 grids at once without human interference. The operation of the system at the user level is solely controlled within the MATLAB environment: the commands to perform sample handling (loading/unloading in the microscope), microscope steering (magnification, focus, image acquisition, etc.) as well as automatic crystal detection have been implemented. Different types of thin samples can efficiently be screened provided that the particular detection algorithm is adapted to the specific task. Hence, operating time can be shared between multiple users. This is a major step towards the integration of transmission electron microscopy into a high throughput work-flow.

Automatic TEM image analysis of membranes for 2D crystal detection.

TEM image processing tools are devised for the assessment of 2D-crystallization experiments. The algorithms search for the presence and assess the quality of crystalline membranes. The retained scenario emulates the decisions of a microscopist in selecting targets and assessing the sample. Crystallinity is automatically assessed through the diffraction patterns of high magnification images acquired on pertinent regions selected at lower magnifications. Further algorithms have been developed for membrane characterization. Tests on images of different samples, acquired on different microscopes led to good results.

Visually guided movements: learning with modular neural maps in robotics.

Robotics involves complex processing and requires modular controllers. For the connectionist approach, the adaptation of each module within the global system remains a major problem to be solved. This paper proposes the idea that biological learning can take advantage of the structures of the modules and the nature of modular decomposition. Therefore, we address this problem starting with the architecture of the system. We illustrate this approach using a robotic application: the visual servoing of the arm's end-effector. The on-line adaptation of a simple controller permits excellent results. To process several variables, and to limit the size of the memory required, this controller is decomposed into modules, in the image of sensorial or motor processing centers. The learning of the modules is realized on-line, a bi-directional architecture permits the adaptation of each module using a simple algorithm. The results obtained with various modular arrangements, both during intensive computer simulations and on our robotic platform, confirm the practical interest of this approach.