LAB TECHNIQUES · With robotic equipment, Harvard team scans cells faster than researchers can blink their eyes
BY MARK D. UEHLING
January 15, 2008 | Whatever its marvels, the human eye has a critical flaw. It's connected to a much larger organ, and that gray heap can only see so much so fast. Fortunately, Steven Altschuler, from Harvard's Bauer Center for Genomics Research, and colleagues have speeded up the process, adapting automated microscopy for high-throughput drug screening.
The approach relies on "hypothesis-free molecular cytology that provides multidimensional single-cell phenotypic information yet is simple and inexpensive enough to allow extensive dose-response profiles for many drugs," Altschuler and colleagues wrote in a paper in Science last November.
The team tested 100 drug compounds in 16 concentrations in hamster cells, arrayed in 386-well plates that were digitally photographed and analyzed according to 93 different variables. The result: 600,000 images and a billion data points occupying 2 terabytes of disk space. This sort of thing may not be possible at a community college but seems feasible at any major campus or company. Spending only $100,000, the Harvard group used a Nikon fluorescence microscope, a cooled camera from Hamamatsu, and a plate-transfer crane from Hudson. It was all controlled by Metamorph software.
Co-author Timothy Mitchison, deputy chair of Harvard's Department of Systems Biology, says the prodigious volume of data was "equal to the number of measurements we'd collected in my lab, ever, times 107." Mitchison could only give the mountain of data to his statisticians, who adapted the Kolmogorov-Smirnov algorithm to detect patterns in the cellular images.
The take-home message is that informatics (in this case, Matlab software, running on a 50-node Linux cluster) can discern both expected visual patterns for similar drugs and unexpected ways in which ostensibly similar drugs generate different heat map profiles. The technique could help industry categorize new compounds with known mechanisms of action — or discover altogether new ways that drugs work. "We can screen 5,000 wells a day," Mitchison says, "which is slow by high-throughput standards but fast by cell biology standards."
Perlman Z.E., et al. "Multidimensional drug profiling by automated microscopy." Science 306, 1194-98; 2004.