CAMBRIDGE, Mass.—High-content screening (HCS) is a data-intensive field and it is often difficult for researchers to discern biologically relevant information within the data morass. To deal with this challenge, researchers at the Novartis Institutes for BioMedical Research and Harvard Medical School recently applied traditional factor analysis to the mining of HCS data. The results were published in Nature Chemical Biology.
The researchers started with the supposition that following treatment with different compounds, groups of image-based cell features showing highly correlated changes between cells are probably describing the same phenotypic properties. Using fluorescent probes for DNA, mitosis, and DNA replication, they designed an HCS assay to identify compounds that impact cell proliferation, monitoring 36 nuclear features. They then used a common factor model to reduce the dimensional space to 6 orthogonal factors: nuclear size; DNA replication; mitosis/chromosome condensation; nuclear morphology; EdU incorporation morphology; and nuclear shape.
In a proof-of-principle study, the researchers used the assay to screen >6500 compounds, including compounds derived from a diversity library, a natural products library, and a collection of known bioactive compounds. They identified 211 hits—most coming from the bioactive library. They examined the hits to define chemical structure and identify known and predicted molecular targets. Surprisingly, they found that the phenotypes correlated better with predicted compound targets than compound structures.
If widely applicable, the HCS data generation and analysis system should provide information-rich compound SARs for researchers looking to better understand how these compounds work in biological systems.
"Dealing with complexities of predictive toxicology will require breakthroughs in cytological image analysis, target prediction schemes, and data mining," the researchers wrote. "Our integration of image-based cytological phenotypes with chemical structure and computational ligand-target prediction represents a step forward in solving this and other difficult drug discovery problems."