Screening Cellular Features for Image-Based Assays
Author Information
Author(s): Trask O. Joseph Jr., Davies Anthony, Haney Steven, Logan David J., Carpenter Anne E.
Primary Institution: The Broad Institute of MIT and Harvard
Hypothesis
A screening approach to choose image-based assay features might be less time, resource, and expertise consuming in at least some situations.
Conclusion
The screening approach is helpful for identifying features suited to score a particular assay when machine learning is not preferred.
Supporting Evidence
- The study demonstrates the use of a screening approach to select the best measures for scoring image sets.
- CellProfiler allows for the measurement of a large number of features, aiding in assay development.
- The Z' factor and V factor were used to assess assay quality, with values above 0.5 considered excellent.
Takeaway
This study shows a way to pick the best measurements for experiments that look at cells using pictures, making it easier and faster to find what works.
Methodology
The study used the open-source software CellProfiler to analyze publicly available image sets and measure various cellular features.
Potential Biases
There is a risk of selecting features that are effective for distinguishing controls based on irrelevant morphological properties.
Limitations
Selecting the best measure from a large library can uncover spurious differences, and care must be taken to avoid selecting features that do not reflect the true phenotype of interest.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website