High-throughput screening (HTS) has been a blessing and a curse. A blessing in codifying a straightforward approach to discovering drug candidates once a particular target has been identified. But a curse in terms of its overall efficiency, with many extremely expensive screens yielding drug candidates that progress to an even more expensive development process, only to fail at later stages.
Each of those failures have taught lessons about specific components in the drug screening process. Turns out, except for the robots, pretty much everything else needed to be looked at carefully. Recall that the concept of HTS is to put one compound into each isolated assay cell and find the ones with activity against the assay. Well, as expected, the target validation and assay design is critical: You get out what you put in, which isn’t always what you expected. For example, many HTS were designed to put in the enzyme and its substrate and read out substrate modification. If the enzyme did something different in vivo than in vitro, there might be a surprise down the line. Similarly, the substrate could be complex in cells, and the peptide fragment on which the enzyme acts could behave very differently as a free isolated entity than in its whole, native protein assembly.
But that didn’t compare to the oft overlooked pitfalls of the chemical library being screened. In addition to compositional flaws, chemical “con men” turned out to be frequent residents of these libraries. These compounds, known as Pan-Assay INterference compounds (PAINs) had a tendency to give a positive readout in a variety of assays and even in computer screening/docking (virtual library screening). This meant that many HTS appeared to have worked and resources were then invested in the later drug development stages, only to find out that the activity was actually non-specific. The more common mechanisms of this phenomenon were aggregation of the compounds into bioactive complexes or highly reactive bonds or chemical groups resulting in covalent attachment to proteins.
The chemical features underlying this false positive behavior were analyzed carefully and published some years ago. Turns out, just about everyone was using a library with these bad actors in them. The report of easy-to-use filters to get rid of these compounds from chemical libraries represented a sea change in drug discovery screening. All hail the dedicated scientists who do the systematic grunt work like this to advance the field.