High-Throughput Screening to Predict Chemical-Assay Interference

Abstract

The U.S. federal consortium on toxicology in the 21 st century (Tox21) produces quantitative, highthroughput screening (HTS) data on thousands of chemicals across a wide range of assays covering critical biological targets and cellular pathways. Many of these assays, and those used in other in vitro screening programs, rely on luciferase and fluorescence-based readouts that can be susceptible to signal interference by certain chemical structures resulting in false positive outcomes. Included in the Tox21 portfolio are assays specifically designed to measure interference in the form of luciferase inhibition and autofluorescence via multiple wavelengths (red, blue, and green) and under various conditions (cell-free and cell-based, two cell types). Out of 8,305 chemicals tested in the Tox21 interference assays, percent actives ranged from 0.5% (red autofluorescence) to 9.9% (luciferase inhibition). Self-organizing maps and hierarchical clustering were used to relate chemical structural clusters to interference activity profiles. Multiple machine learning algorithms were applied to predict assay interference based on molecular descriptors and chemical properties. the best performing predictive models (accuracies of ~80%) have been included in a web-based tool called InterPred that will allow users to predict the likelihood of assay interference for any new chemical structure and thus increase confidence in HTS data by decreasing false positive testing results. Chemical hazard assessment testing in the twenty-first century has evolved to encompass large high-throughput screening (HTS) research programs, designed to produce quantitative data on the activity of thousands of chemicals across hundreds of biological targets and pathways, a strategy that has long been used in drug discovery. Such efforts, exemplified by the federal Tox21 research consortium 1,2 , are intended to facilitate rapid chemical hazard screening, predictive computational toxicology using machine learning and artificial intelligence techniques, and human-relevant systems biology models that provide mechanistic insight into chemical toxicity (e.g. 3). The Tox21 program and other such HTS initiatives rely upon an array of biological assays, i.e. analytical measurement procedures defined by a set of reagents that produce a detectable signal, allowing a biological process to be quantified 4. Many HTS platforms use cell-based assays measuring processes such as cell growth/death, receptor binding, or protein expression while others are cell-free assays that characterize biochemical activity. Both formats use a variety of detection technologies including fluorescence and luminescence readouts. Fluorescence-based assays are one of the leading technologies in terms of widespread application reported in PubChem 5,6. Indeed, the routinely used fluorescence intensity-based format allows for optimization of speed, accuracy, reproducibility and sensitivity of assays 7. Luminescence is frequently used as a readout from luciferase-based reporter gene assays and pro...

Development of a Microscale Thermophoresis-Based Method for Screening and Characterizing Inhibitors of the Methyl-Lysine Reader Protein MRG15

“…This choice relies on the evidence that many screening compounds and reference molecules interfere at UV and blue wavelengths. 30,31 In particular, we selected the NanoTemper proprietary second-generation red dye (Monolith Protein Labeling Kit RED-NHS 2nd Generation, NanoTemper Technologies) that carries a reactive NHS-ester group that reacts with primary amines of lysine residues to form a covalent bond. It is reported that the second-generation dye gives enhanced binding amplitudes and higher signal-to-noise (S/N) ratios with respect to the first-generation dye.…”

Evaluation of Firefly and Renilla Luciferase Inhibition in Reporter-Gene Assays: A Case of Isoflavonoids

“…InterPred webserver (version 1.0, accessed on 3 August 2020, National Institute of Environmental Health Sciences, Durham, NC, USA) is a platform for the prediction of interference in cellular and biochemical assays that use fluorescence or luminescence as the detection technology [23]. InterPred includes 17 QSAR models that were developed using machine learning, and are based on data from more than 8,000 unique structures in assays contained in the Tox21 Consortium library, i.e., luciferase inhibition assays (tox21luc-biochem-p1), and autofluorescence assays in the HepG2 and HEK-293 cell lines, and the cell-free assay formats (tox21-spec-hepg2-p1, tox21-spec-hek293-p1, respectively) [1,23]. The intent of InterPred is to enable the identification of false-positive results due to autofluorescence and inhibition of luciferase in large high-throughput screening assays.…”

“…Advances in high-throughput screening of compounds for biological activity have enabled the compilation of data from cell-based and biochemical assays on a myriad of biological targets for thousands of compounds, such as the US Federal Consortium on Toxicology in the 21st Century (Tox21 Consortium) program [1,2]. In this way, our knowledge of the biological activities of compounds has expanded rapidly, whereby the applicability of a compound (e.g., as a drug) and its potential toxicity can be identified early in the process of drug discovery.…”

“…Biological activities are most often detected in assays using fluorescence (as~53% of cases), followed by luminescence (~21%), and absorbance (<7%) [3]. The interpretation of the data obtained can often, however, be challenging, due to interference in the assays, such as quenching and autofluorescence in fluorescence-based assays [1], inhibition and stabilization of the enzyme in luminescence-based assays [3][4][5], and interference of colored compounds in absorbance-based assays [6]. All such situations can lead to false-positive results.…”

“…Predicting potential inhibition of FLuc (and hence the potential for false-positive results in reporter-gene assays) using different chemoinformatic approaches has been a relatively active field in recent years. This has led to the development of different freely available models that can be used when designing studies using FLuc as the reporter protein, such as Luciferase Advisor [21], ChemFLuc [22], and InterPred [1,23]. We were, however, not able to find any in silico models to predict the potential for inhibition of RLuc.…”