Integration of Multiple Readouts into the Z′ Factor for Assay Quality Assessment

Abstract

Methods that monitor the quality of a biological assay (i.e., its ability to discriminate between positive and negative controls) are essential for the development of robust assays. In screening, the most commonly used parameter for monitoring assay quality is the Z' factor, which is based on 1 selected readout. However, biological assays are able to monitor multiple readouts. For example, novel multiparametric screening technologies such as high-content screening provide information-rich data sets with multiple readouts on a compound's effect. Still, assay quality is commonly assessed by the Z' factor based on a single selected readout. This report suggests an extension of the Z' factor, which integrates multiple readouts for assay quality assessment. Using linear projections, multiple readouts are condensed to a single parameter, based on which the assay quality is monitored. The authors illustrate and evaluate this approach using simulated data and real-world data from a highcontent screen. The suggested approach is applicable during assay development, to optimize the image analysis, as well as during screening to monitor assay robustness. Furthermore, data sets from high-content imaging assays and other state-of-the-art multiparametric screening technologies, such as flow cytometry or transcript analysis, could be analyzed.

Kernelized Z’ factor in multiparametric screening technology

“…In terms of new vector α the objective J (α) becomes, (12) Correspondingly, a pattern in the original input space R n is mapped into a potentially much higher dimensional feature vector in the feature space H. The scatter matrices in kernel space can expressed in terms of the kernel only as follows: (13) (14) (15) (16) Popular choice is the Gaussian kernel where S B is the "between positive and negative control scatter matrix", S W is the "within controls scatter matrix" and w T is a vector transpose. Note that due to the fact that scatter matrices are proportional to the covariance matrices we could have defined J using covariance matrices-the proportionality constant would have no effect on the solution.…”

“…At present, there is multivariate Z' factor to monitor assay quality based on multiple readouts simultaneously. 12 Such a method enables the assessment of the image readouts' suitability to monitor relevant biological effects.…”

SLAS Europe High-Content Screening Conference in Dresden: A Glimpse of the Future?

“…Kerz et al 2 in this collection use the first component of their principal component analysis of time series to follow the evolution of their cells over time. A multiparametric extension of the Z′ quality control criteria has also been proposed by Kümmel et al, 16 and the Mahalanobis distance has been used for quantifying phenotypic strength. 17 The problem with these tools is that they do not capture the diversity of phenotypes found in the assays.…”

Image-Based High-Content Screening in Drug Discovery

“…To explore the high content of images through multi-parametric data analysis more sophisticated tools are needed. The combination of several parameters can improve the assay quality, however, the statistical parameter, Z' value, which is commonly used for single readouts needs to be adapted for the analysis of multiple features (Kümmel et al, 2010). In order to enable the analysis of dozens of features in parallel we have generated an inhouse software package (Kümmel et al, 2011).…”