Of the many drug candidates entering clinical trials, less than 10 percent make it to market—often because the drug is toxic or does not work. Getting drug candidates through clinical trials is costly and time consuming, so it is crucial to find ways to improve the quality of pre-clinical drug screens.
Drug screening is an essential part of the drug development process that sifts through a large library of compounds, picking out those that are active against a target structure. Up until the 1980s, drug screening was a lengthy, low-throughput process, where the most tech-heavy laboratories screened no more than 100 compounds a week. With advanced technologies and integration with automated systems, laboratories can now screen upwards of 100,000 compounds a day. Throughput is no longer the major limiting factor in drug screening; focus now lies on how to enhance the quality of these automated drug screening systems.
What are high-throughput drug screening systems?
High-throughput drug screening systems typically take a target-based or phenotypic approach. Target-based drug screening relies on the compound library being screened against purified and isolated target structures (usually proteins), revealing drug–target binding interactions. Phenotypic drug screening relies on the compound library being screened against cells, observing drug responses through phenotypic changes in the cells.
Target-based drug screening reveals specific drug–target interactions but misses the complex environment of a cell, meaning these studies can lack information on factors such as permeability, distribution, and off-target interactions. Consequently, the toxicity and ability of a drug to work in a living system may not be fully elucidated until it reaches clinical trials. Scientists are now looking to modern forms of phenotypic drug screening to improve the quality of drug screens. Here, quality refers to screening information being robust and translatable to human use, ideally targeting the biological mechanism for a given pathology without toxicity.
In phenotypic drug screening, automated imaging systems reveal multiple drug features (such as distribution and permeability) simultaneously. Traditionally, high-throughput imaging systems measured one or two features, usually through fluorescent signals, in multiwell plate formats. But these simplified systems average the signal over the whole well, losing drug-response differences between cells—a problem for pathologies with heterogeneous cell populations (such as cancer), where individual cells can have different responses to a drug.
Advances in microscopy and image analysis led to high-content screening, where multiple streams of imaging data are simultaneously quantified. Through the combination of sophisticated microscopy equipment and AI-based analysis software, automated high-content imaging systems reduce human bias and handle large datasets to analyze multiple features of cell populations. Consequently, high-content screening techniques are capable of measuring drug response differences in heterogeneous cell samples.
Drug screening culture strategies to improve quality
In vivo drug screening methods can improve the quality of drug screening studies, but it is not feasible to run high-throughput screening studies in animals. High-throughput experiments in animals pose ethical challenges and high costs, and animals do not necessarily reflect the pathophysiology of human disease.
High-throughput drug screening experiments often use cells cultured in multiwell plates with cells grown a single layer thick. Multiwell plates allow high-throughput experiments to be performed as plates easily integrate with automated systems. These ”2D” cell cultures are a valuable tool in research, but with so many drug candidates failing at clinical trials, consideration should be given to how well these cultures reflect the tissue of a target pathology in drug screening experiments.
Researchers are examining more complex in vitro culture systems, such as 3D cultures, in hope of improving the quality of drug screens. 3D culture systems aim to better reflect in vivo biological conditions. These cultures are designed so cells can grow in all directions (unlike 2D cultures) while making attachments with extracellular matrix components and other cells that more closely simulate the complex tissue environment cells experience in vivo.
As 3D culture techniques developed, more options became available for use in high-throughput screening, from partly-automated drug screening systems using hydrogel-based 3D cultures in 384-well plates to fully-automated drug screening systems using complex 3D cultures (organoids) in 96-well plates. Recently, anti-cancer drugs have been screened against organoids grown from patient-derived cancer cells. By using patient-derived cells, the drug screening process can capture different responses to therapies that arise from heterogeneous cancer cells (both within a tumor and between patients), showing its potential in patient-tailored drug development. Organoids can also be combined with other types of cells, such as immune cells, which could build a more detailed picture of disease pathology.
Developing robust drug screening methods to increase quality
Quality drug screens rely on screening data being reliable, reproducible, and valid. With laboratories using different assay formats, detection methods, and automation systems, variability can exist between study conclusions. Researchers have looked at drug screening assay development and statistical methods to generate more robust methods. Shockley et al. recently developed “CASANOVA” (cluster analysis by subgroups by analysis of variance) to improve reproducibility by presenting a method of standardizing high-throughput results. This statistical approach allows variable and consistent drug responses to be identified between datasets, offering a method of standardizing drug screening studies between different laboratories.
Many high-throughput screening techniques (both target-based and phenotypic) use fluorescent probes (dyes or labelled proteins/antibodies) for measuring drug responses and interactions. These probes must be carefully selected to avoid false positive and negative results arising from factors such as autofluorescence, quenching, or off-target interactions. But compounds can be difficult to label and probes can interfere with drug binding and cell functioning. Recently, Kobayashi et al. presented a label-free approach using machine learning to identify morphological changes after drug treatment in bright-field images. Further development in label-free drug screening techniques may overcome some of the limitations encountered when using fluorescent labels or probes, opening up automated systems that are linked to other detection methods (such as mass spectrometry).
Considering the costs of quality drug screens
Automated systems have amplified the number of compounds that laboratories can screen. Automation is not only making drug screening faster but also more sophisticated, with imaging technologies available that gather multiple streams of drug response information from individual cells. While these advances improve the quality of drug screens, the equipment can be expensive and require training and maintenance—hiking up the prices of drug screening studies. With automated systems varying from robotic arms that move plates to fully integrated workflows that handle liquids and acquire and analyze data, laboratories manage these costs by tailoring the extent of automated systems based on their needs and budget. But with accelerating use and development of high-throughput systems, plus cost savers like open-source software and 3D-printed parts, automated drug screening aims to become more affordable.