Automated quality control tool for high-content imaging data by building 2D prediction intervals on reference...

Area of Science:

• Cellular imaging
• High-content screening
• Drug discovery

Background:

• Automated microscopy and image analysis, including Cell Painting, enable large-scale quantitative profiling of cellular phenotypes.
• Phenotypic profiles are crucial for studying chemical perturbations and identifying on- and off-target effects during drug lead optimization.
• Maintaining consistent quality of Cell Painting assays is essential for building reliable phenotypic databases.

Purpose of the Study:

• To introduce an automated tool for assessing and controlling the quality of Cell Painting assays over time.
• To quantify the reproducibility of cellular biosignatures from annotated reference compounds.
• To establish a mechanism for detecting aberrations in new Cell Painting experiments.

Main Methods:

• The tool learns historical biosignatures of reference treatments.
• It constructs a two-dimensional probabilistic quality control (QC) limit based on learned biosignatures.
• The QC limit is applied to new Cell Painting experiments to detect deviations.

Main Results:

• The tool was validated using simulated data and demonstrated on Cell Painting data from the A549 cell line.
• It provides a sensitive and detailed mechanism for quality assessment.
• The system is designed to be easy to interpret for users.

Conclusions:

• The developed automated tool addresses the need for quality control in Cell Painting assays.
• It enables reliable validation of assay quality by monitoring biosignature reproducibility.
• This facilitates more robust and trustworthy large-scale phenotypic profiling in drug discovery.