Repurposing High-Throughput Image Assays Enables Biological Activity Prediction for Drug Discovery

Abstract

In both academia and the pharmaceutical industry, large-scale assays for drug discovery are expensive and often impractical, particularly for the increasingly important physiologically relevant model systems that require primary cells, organoids, whole organisms, or expensive or rare reagents. We hypothesized that data from a single high-throughput imaging assay can be repurposed to predict the biological activity of compounds in other assays, even those targeting alternate pathways or biological processes. Indeed, quantitative information extracted from a three-channel microscopy-based screen for glucocorticoid receptor translocation was able to predict assay-specific biological activity in two ongoing drug discovery projects. In these projects, repurposing increased hit rates by 50- to 250-fold over that of the initial project assays while increasing the chemical structure diversity of the hits. Our results suggest that data from high-content screens are a rich source of information that can be used to predict and replace customized biological assays.

Classifying T cell activity in autofluorescence intensity images with convolutional neural networks

“…Except for the trivial frequency classifier, all models can be represented as a form of neural network with different input features and architectures. Also, we select features that are easy to extract from cell images, such as the raw pixel matrix and total intensity, as well as Cel-lProfiler attributes, which are commonly used in cellular image classification studies [8,37].…”

Next-Generation Phenotypic Screening in Early Drug Discovery for Infectious Diseases

“…Profiles of cell populations treated with different experimental perturbations can be compared in order to suit many goals, such as identifying the phenotypic impact of chemical or genetic perturbations, grouping (clustering) compounds and/or genes into functional pathways, and identifying signatures of disease [72]. More recently, this idea has been extended further: Ceulemans et al [73] demonstrated that 'image-based fingerprints' of compounds derived from a given image-based cellular assay can be repurposed to predict the biological activity of those same compounds in other seemingly unrelated assays, even those targeting alternate pathways or biological processes. The approach has the potential to greatly reduce unnecessarily voluminous scaling of screens by predicting the likelihood of activity on a new target.…”

“…Focusing on the cardiomyocyte model, the same teams most recently went on to report the use of molecular phenotyping as a means to augment high-content image-based drug screening by helping to assure and stratify lead compound selection and characterize MoA classes [76]. [68,73]). Set inside the red arrow, left to right, is a schematic representation workflow showing steps: (1) chemical phenotypic screening that used a three to five channel readout, N500 000 compounds, against several distinct cell lines; (2) image analysis, N-dimensional feature space extrapolating N800 independent feature parameters per single cell; and (3) cell profile fingerprint yielding a large array of single-cell morphological feature data.…”

“…Set inside the red arrow, left to right, is a schematic representation workflow showing steps: (1) chemical phenotypic screening that used a three to five channel readout, N500 000 compounds, against several distinct cell lines; (2) image analysis, N-dimensional feature space extrapolating N800 independent feature parameters per single cell; and (3) cell profile fingerprint yielding a large array of single-cell morphological feature data. In the context of 1200 previously performed bioactivity assays applicable to the compound library used by Simm et al [73], this shows that machine-learning methods can be applied to yield useful enhanced predictivity for these same compounds in new assays [73]. (B) Molecular phenotyping (adapted from [75,76]), set inside the blue arrow, showing a schematic illustration of the workflow for molecular phenotyping.…”

Tales of 1,008 Small Molecules: Phenomic Profiling through Live-cell Imaging in a Panel of Reporter Cell Lines

“…In the last two decades, technological and analytical innovations in high-throughput microscopy and transcriptomics have enabled the large-scale phenomic profiling of small-molecule libraries [1][2][3][4][5][6][7][8]. These profiling approaches quantify the phenotypic response of cells to compound treatment by simultaneously measuring changes in hundreds or thousands of features, be they transcript levels assessed with gene expression technologies or the morphological characteristics of cells in a microscopy image.…”