Single-Cell Level Estrogen Receptor Activity Data

This dataset contains normalized single-cell level data for 60 reference compounds analyzed with high throughput microscopy and high content analysis-based experiments that were performed using the GFP-ER⍺:PRL-HeLa cell line. The multidimensional imaging data was used to train a classification model to ultimately predict the impact of unknown compounds on the estrogen receptor, either as agonists or antagonists as outlined in the RMarkdown file. All chemical screening data were normalized to E2-treated control samples using the robust z-score calculation based on single-cell data. In experiments that spanned multiple sample plates, the median and MAD values were calculated for each plate, the median MAD value across all plates was determined, and the robust z-score was calculated using the per-plate median value and the overall median MAD value. Datasets with names \"LMH10\" and \"M50\" are engineered datasets with number of features expanded using percentiles. Percentiles were determined by identifying the single-cell features falling into the 0-10th percentile range (L10), 45th–55th percentile range (M10), the 90th–100th percentile range (H10), and the 25th–75th percentile range (M50) from each sample. Experimental data sets containing all single-cell data and engineered data sets containing only the L10 + M10 + H10 or only the M50 single-cell features were also divided into 2 subsets depending on the detection of visible GFP-ER nuclear spots (all cells and array positive cells) used for subsequent model development.