By Salvatore Salamone
May 29, 2008 | In high-throughput screening processes used for drug discovery, tens of thousands of wells each containing tens or hundreds of cells, need to be analyzed each day. Automated analysis of the cellular relationships within so many wells is usually limited by a number of factors.
“Current automated screening systems for examining cell cultures look at individual cells and do not fully consider the relationships between neighboring cells,” said Geoffrey Gordon, associate research professor in the Carnegie Mellon University's School of Computer Science's Machine Learning Department. “This is in large part because simultaneously examining many cells with existing methods requires impractical amounts of computational time.”
To address this issue, Gordon, biomedical engineering student Shann-Ching Chen and computational biologist Robert F. Murphy (of the university’s Lane Center for Computational Biology) have applied an artificial intelligence (AI) technique to cellular image analysis that has the potential to significantly speed up critical steps in an automated method for analyzing cell cultures and other biological specimens.
The researchers published their findings in a paper last month (“Graphical Models for Structured Classification, with an Application to Interpreting Images of Protein Subcellular Location Patterns,” Journal of Machine Learning Research, Volume 9, p. 651 to 682, April 2008).
In their paper, they discussed the benefits of using current computer vision systems to distinguish patterns that are difficult for humans to detect. But they noted how in many cases the automated systems may confuse similar patterns. This confusion can only be resolved by considering neighboring cells – something that has been impractical because of computing limitations.
That’s where the new technique comes in. The CMU researchers were able to expand the analysis to multiple cells by increasing the efficiency of the belief propagation algorithm, which is an algorithm widely used in many artificial intelligence applications.
The researchers analyzed protein patterns within HeLa cells and found their technique accelerated analysis by several orders of magnitude.
Broader application of this approach has great potential. Its “improved accuracy could reduce the cost and the time necessary for these screening methods, make possible new types of experiments that previously would have required an infeasible amount of resources, and perhaps uncover interesting but subtle anomalies that otherwise would go undetected,” the researchers said.