High throughput image analysis on PetaFLOPS systems

Abstract

Today's state of the art high-throughput screening facilities can produce tens of thousands of images of cells per day. Analyzing images from high-throughput screening experiments is very time consuming and computationally demanding. Researchers are currently limited not by the availability of experimental data, but by the computing resources for the image analysis. The Max Planck Institute of Molecular Cell Biology and Genetics Dresden, Germany, (MPI-CBG) and the Center for Information Services and High Performance Computing at the Technische Universität Dresden (ZIH) are working together to integrate high performance computing systems into the workflow of biologists. The MPI-CBG has developed software that biologists use for their image analysis work. The software can utilize local workstations and remote HPC systems for image analysis. Currently the software is used successfully on small clusters and PC-Farms. Most parts of the image analysis workflow of screening experiments can be performed in parallel and is ideal for distribution on large systems. With a few modifications and a new approach to data management, the software should be able to scale to PetaFLOPS systems. © Springer-Verlag Berlin Heidelberg 2007.