Impact of normalization methods on high-throughput screening data with high hit rates and drug testing with dose–response data

Abstract

Motivation: Most data analysis tools for high-throughput screening (HTS) seek to uncover interesting hits for further analysis. They typically assume a low hit rate per plate. Hit rates can be dramatically higher in secondary screening, RNAi screening and in drug sensitivity testing using biologically active drugs. In particular, drug sensitivity testing on primary cells is often based on dose–response experiments, which pose a more stringent requirement for data quality and for intra- and inter-plate variation. Here, we compared common plate normalization and noise-reduction methods, including the B-score and the Loess a local polynomial fit method under high hit-rate scenarios of drug sensitivity testing. We generated simulated 384-well plate HTS datasets, each with 71 plates having a range of 20 (5%) to 160 (42%) hits per plate, with controls placed either at the edge of the plates or in a scattered configuration.Results: We identified 20% (77/384) as the critical hit-rate after which the normalizations started to perform poorly. Results from real drug testing experiments supported this estimation. In particular, the B-score resulted in incorrect normalization of high hit-rate plates, leading to poor data quality, which could be attributed to its dependency on the median polish algorithm. We conclude that a combination of a scattered layout of controls per plate and normalization using a polynomial least squares fit method, such as Loess helps to reduce column, row and edge effects in HTS experiments with high hit-rates and is optimal for generating accurate dose–response curves.Contact: john.mpindi@helsinki.fiAvailability and implementation, Supplementary information: R code and Supplementary data are available at Bioinformatics online.

High-throughput organoid screening enables engineering of intestinal epithelial composition

“…Results (excel or .csv files) were converted into a data frame containing raw assay measurements corresponding to metadata for plate position, treatments, doses, cell type, and stimulation. Raw values were log10 transformed, then a LOESS normalization was applied to each plate and assay to remove systematic error and column/row/edge effects using the formula (Mpindi et al, 2015):…”

Time-resolved mapping of genetic interactions to model rewiring of signaling pathways

“…Per feature, the feature's minimum value was added to each value prior to logn(x+1) transformation to approach the features' histogram to normal distribution. Following transformation, each plate in each screen was normalized separately for each feature by B-Score normalization (Ljosa et al, 2013;Mpindi et al, 2015). The B-Score normalization centers and scales the data to be the residuals of the median polish divided by the median absolute deviation (mad) across all values of the plate and thus be symmetrically centered around zero and scaled in units of the mad.…”

High Content, Phenotypic Assays and Screens for Compounds Modulating Cellular Processes in Primary Neurons

“…Besides normalization, the B score method uses a two-way median polish algorithm to minimize within-plate column and row effects (Brideau, Gunter, Pikounis, & Liaw, 2003). Other noise reduction methods based on Tukey median polish algorithm or using polynomial least squares fit have also been developed to correct for systematic within plate effects (Makarenkov et al, 2007; Mpindi et al, 2015). The Loess local polynomial fit is based on the least squares polynomial approximation, and provides more reliable data and less false positives on plates with higher hit rates than 20% than the B score method, which could be attributed to B score’s dependency on the median polish algorithm (Mpindi et al, 2015).…”

“…Other noise reduction methods based on Tukey median polish algorithm or using polynomial least squares fit have also been developed to correct for systematic within plate effects (Makarenkov et al, 2007; Mpindi et al, 2015). The Loess local polynomial fit is based on the least squares polynomial approximation, and provides more reliable data and less false positives on plates with higher hit rates than 20% than the B score method, which could be attributed to B score’s dependency on the median polish algorithm (Mpindi et al, 2015). Combining a polynomial least squares fit method with normalization of data to scattered controls may provide the best option for screens with high hit rates and also for generating accurate dose-response curves.…”